Preparing a Training Image

Obtaining a Training Image

Use either of the following methods to obtain a training image:

• (Recommended) Download the training base image matching the driver version from Ascend Hub based on the system architecture (Arm or x86) and model framework (TensorFlow, PyTorch, or MindSpore). Change the default user in the container to root based on the base training image. (In versions later than 21.0.4, the default user of the base training image is non-root.) The base image does not contain files such as the training script and code. During training, those files are mapped to the container.
• Customize your own training image by referring to Common Operations.

You can rename the downloaded or created training image, for example, trainning:v3.0.0.

(Recommended) Hardening the Image

For details, see Hardening Container Security.