Help Center/ ModelArts/ Image Management/ Using a Custom Image
Updated on 2026-08-27 GMT+08:00
View PDF
Share

Using a Custom Image

If the preset ModelArts images do not meet your needs, such as requiring a specific version of a deep learning framework, system library, or service dependency, you can create a custom image and upload it to SWR. You can then use this image for ModelArts model training, real-time inference, and notebook instances.

Prerequisites

A custom image has been created. For details, see Creating a Custom Image.

Billing

Step 1: Pushing a Custom Image to SWR

  1. Log in to the SWR console and push the custom image to the SWR image repository.
  2. In the navigation pane of the SWR console, choose My Images. Refresh the page to view the new image. Click the image name to view the image details (including the image version and size). Ensure that the image details are the same as those of the pushed image.
    Figure 1 My image details

Step 2: (Optional) Sharing a Custom Image with Other Accounts

  • SWR basic edition: After pushing a private image, you can share it with other users and grant them permission to pull it. The users with whom the image is shared can log in to the SWR console and view the shared image on the My Images > Shared Images page. For details, see Sharing a Private Image.
  • SWR enterprise edition: If you have multiple accounts and want to use the same repository, you can use Resource Access Manager (RAM) to share the repository among the accounts. For details, see Sharing a Repository.

Step 3: (Optional) Registering a Custom Image

You can directly use SWR images for model training and real-time inference without registering them. You can also use images that have been registered with ModelArts. In the notebook scenario, you must register an image with ModelArts before you can select the image when creating a notebook instance. For details, see Registering an Image.

Step 4: Using a Custom Image

You can use custom images for model training, real-time inference, and notebook instances.

  • New console: When creating a custom training job, select a custom image in the Training Configuration section and click . In the Image panel, select an image as required. For details, see Creating a Custom Training Job.
    Figure 2 Training Configuration > Select Image

    Table 1 Model Training > Select Image

    Image Type

    Description

    SWR
    • My Images: image pushed to SWR basic edition in step 1.
    • Shared Images: images shared by others in step 2.

    SWR Enterprise Edition

    Image pushed to SWR enterprise edition in step 1.

    Registered Images

    Image registered with ModelArts image management in step 3.
  • Old console: When creating a training job, set Boot Mode to Custom image in Environment Settings, click , and select an image on the Image panel. For details, see Creating a Production Training Job.
    Figure 3 Environment Settings > Boot Mode

When deploying a real-time service, select an image in any of the following ways in the Unit Settings area: For details, see Inference Deployment.

  • Method 1: Choose a custom image.
  • Method 2: Choose the image URL option and enter the SWR image path.
    Table 2 Real-Time Inference > Select Image

    Image Type

    Description

    SWR
    • My Images: image pushed to SWR basic edition in step 1.
    • Shared Images: images shared by others in step 2.

    SWR Enterprise Edition

    Image pushed to SWR enterprise edition in step 1.

    Registered Images

    Image registered with ModelArts image management in step 3.
When creating a notebook instance, select Custom Images in the Environment section and click . In the Registered Images tab of the Image panel, select the image registered in step 3. For details, see Creating a Notebook Instance.
Figure 4 Environment > Custom Images