Application Scenarios and Process

If preset images cannot meet your service requirements, you can create container images based on the preset images for development and training.

Generally, you will need to reconstruct the ModelArts development environment, for example, by installing, upgrading, or uninstalling some packages. However, the root permission is required when certain packages are installed or upgraded. The running notebook instance does not have the root permission. As a result, you need to install the software that requires the root permission in the notebook instance, which is currently unavailable in the preset development environment.

You need to write a Dockerfile based on a preset public image to customize your image. Then, debug the image so that it can be used in ModelArts. At last, register the image with ModelArts so that it can be used to create development environments to meet your service requirements.

This example shows how to use ma-cli commands in ModelArts CLI to create and register a custom image for AI development with a PyTorch base image. For details, see ma-cli Image Building Command. The following figure shows the whole process.

Figure 1 Creating an image