Documentation

Installation Preparations

  • The host OS meets the compatibility requirements. For details, see Table 2.
  • The driver and firmware have been installed on the host. For details, see NPU Driver and Firmware Installation.
  • You need to install Docker on the host. If the host version is Ubuntu 22.04, the Docker version must be 21.10 or later.
  • Obtain the OS base image.
    1. Run the docker pull {OS name:OS version} command to pull the OS base image. The OS name must be in lower case letter. For example, if the OS base image to be pulled is Ubuntu 20.04, run the docker pull ubuntu:20.04 command.
    2. Run the docker images command to check whether the required container image exists on the host machine.

After the preceding preparations are complete, perform the following steps to deploy the container:

  1. Manually build a CANN container image by referring to the following sections.
  2. Install the container engine plugin by referring to Installing Ascend-Docker.
  3. Use the Ascend Docker plugin to start and deploy the container. For details, see Deploying a Container Image..
  4. (Optional) If you need to set up a container image that contains AI framework training in a container, install the AI framework in the container by referring to the following guide after the preceding deployment is complete.
    • PyTorch: For details about how to install, see Ascend Extension for PyTorch Configuration and Installation.
    • TensorFlow: Install the required version by referring to "Installing TensorFlow 1.15" and "Installing the Framework Plugin Package" in TensorFlow 1.15 Model Porting Guide or "Installing TensorFlow 2.6.5" and "Installing the Framework Plugin Package" in CANN TensorFlow 2.6.5 Model Porting Guide.
    • MindSpore: Install MindSpore of the required version by referring to the MindSpore official website.