openEuler AI Container Image User Guide ​

Introduction ​

The openEuler AI container images package SDKs for different hardware accelerators, along with AI frameworks and large-model applications. You only need to load the image and start a container in the target environment to develop or use AI applications, significantly reducing deployment and environment configuration time and improving efficiency.

Obtain Images ​

Currently, openEuler provides container images for both Ascend and NVIDIA platforms. You can find them here:

• docker.io/openeuler/cann: SDK-type images that install the CANN software stack on top of the openEuler base image, for Ascend environments.
• docker.io/openeuler/cuda: SDK-type images that install the CUDA software stack on top of the openEuler base image, for NVIDIA environments.
• docker.io/openeuler/pytorch: AI framework images that install PyTorch on top of an SDK image; the applicable platform depends on the installed SDK.
• docker.io/openeuler/tensorflow: AI framework images that install TensorFlow on top of an SDK image; the applicable platform depends on the installed SDK.
• docker.io/openeuler/llm: Model application images that include specific large models and toolchains on top of an AI framework image; the applicable platform depends on the installed SDK.

For detailed classifications and tag conventions of AI container images, see oEEP-0014.

Because AI container images are typically large, it is recommended to pull the image to your development environment before starting a container:

docker pull image:tag

Here, image is the repository name, such as openeuler/cann, and tag is the target image tag. After the image is pulled, you can start the container. Note that to use the docker pull command, Docker must be installed as described below.

Start a Container ​

1. Install docker in your environment. Refer to the official guide https://docs.docker.com/engine/install/, or install directly with the following commands:

   sh

   yum install -y docker

   or

   sh

   apt-get install -y docker

2. For NVIDIA environments, install nvidia-container components.

   (1) Configure yum or apt repositories

   • For yum-based systems, run:
   sh

   curl -s -L https://nvidia.github.io/libnvidia-container/stable/rpm/nvidia-container-toolkit.repo | \
   sudo tee /etc/yum.repos.d/nvidia-container-toolkit.repo

   • For apt-based systems, run:
   sh

   curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg

   sh

   curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
   sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
   sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

   (2) Install nvidia-container-toolkit and nvidia-container-runtime:

   sh

   # yum installation
   yum install -y nvidia-container-toolkit nvidia-container-runtime

   sh

   # apt installation
   apt-get install -y nvidia-container-toolkit nvidia-container-runtime

   (3) Configure Docker

   sh

   nvidia-ctk runtime configure --runtime=docker
   systemctl restart docker

   Skip this step if you are not on an NVIDIA platform.

3. Ensure the appropriate driver and firmware are installed. You can obtain the correct versions from the official websites of NVIDIA or Ascend. After installation, test with npu-smi for Ascend platforms or nvidia-smi for NVIDIA platforms. If hardware information is displayed correctly, the installation is successful.

4. After completing the above steps, use the docker run command to start a container.

# Start a container on an Ascend environment
docker run --rm --network host \
           --device /dev/davinci0:/dev/davinci0 \
           --device /dev/davinci_manager --device /dev/devmm_svm --device /dev/hisi_hdc \
           -v /usr/local/dcmi:/usr/local/dcmi -v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
           -v /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/ \
           -ti image:tag

# Start a container on an NVIDIA environment
docker run --gpus all -d -ti image:tag