AI Asset Management of Platform for AI (PAI) provides the custom image feature that allows you import existing images or build new images on the platform. This feature can meet various development and training requirements, improving efficiency and convenience.
Overview
The custom image feature of AI Asset Management allows you to add images based on your business requirements. You can either import existing images or build new images.
Import existing images: You must build images on your on-premises machine or an ECS instance, push them to Alibaba Cloud Container Registry, and then enter the Container Registry image information in the PAI console. This way, the images can be registered as custom images of PAI. This method allows you to use existing images, but the process is relatively complex.
Build new images: You only need to enter the Container Registry image information and configure the job running and image building settings in the PAI console. After the image building job is complete, the image is pushed to Container Registry and registered as a custom image of PAI. This method simplifies the operation process, meeting your needs for one-stop use of custom images and significantly improving efficiency and convenience.
Permission requirements
Alibaba Cloud account: has the permissions to perform all operations.
RAM user: must be added as a workspace member and granted the role permissions. For more information, see Configure Member and Role.
Register an image: import an existing image
Go to Workspaces page, find the workspace that you want to manage, and then click its name. In the left-side navigation pane of the Workspace Details page, choose AI Asset Management > Images. On the Custom Image tab, click Register Image. Then, select Import Existing Image to register your existing image from Container Registry as a custom image of PAI.
The following table describes the key parameters.
Parameter | Description |
Image Type |
|
Enterprise Edition Instance (Container Registry)/Image Namespace/Image Repository/Image Version/Custom Domain Name | Select an Enterprise Edition instance, namespace, repository, image version, and custom domain name, or go to the Container Registry console to create them. Note
|
Visibility |
|
Chip Type | Support the CPU and GPU types to ensure that the registered image can run properly in specific environments. |
Register an image: build an image
Go to Workspaces page, find the workspace that you want to manage, and then click its name. In the left-side navigation pane of the Workspace Details page, choose AI Asset Management > Images. On the Custom Image tab, click Register Image. Then, select Build New Image. After the image is built, the image is pushed to Container Registry and registered as a custom image of PAI.
The following table describes the key parameters.
Parameter | Description |
Basic Information | |
Image Type | Only Enterprise Edition Image (Container Registry) is supported. For more information about Enterprise Edition and Personal Edition images, see What is Container Registry? |
Enterprise Edition Instance (Container Registry)/Image Namespace/Image Repository/Custom Domain Name | Select an Enterprise Edition instance, namespace, repository, image version, and custom domain name, or go to the Container Registry console to create them. Note If the message indicating that you do not have permissions appears, attach the AliyunContainerRegistryReadOnlyAccess policy to your RAM user. |
Visibility |
|
Chip Type | Support the CPU and GPU types to ensure that the registered image can run properly in specific environments. |
Running Information | |
Resources | Select the resources for building the image. We recommend that you choose appropriate resources based on the dependencies in Build Configurations or Dockerfile instructions. For example, if the Dockerfile contains program compilation instructions, the requirement for resources is relatively high. |
VPC Configuration/Security Group | Select the virtual private cloud (VPC) in which the Enterprise Edition instance resides. Note If the image building job requires Internet access, such as the third-party image address and dependency installation in Build Configurations, you must create an Internet NAT gateway, associate an elastic IP address (EIP) with the instance, and configure SNAT. For more information, see Improve Internet access rate by using a private gateway. |
Build Configurations | |
Build Method |
|
View an image building job
On the Custom Image tab, click View Jobs. On the Image Building Job page, find the image building job that you want to manage and click Details in the Actions column.
In the Image Building Job Details panel, click View Logs and View Monitoring to view the logs and monitoring information.
View a custom image
After an image is registered, you can view the registered custom image on the Custom Image tab. Then, you can use the custom image for training later.
References
After you register an image as a custom image of PAI, you can use it in training jobs.