The workspace is a key concept in Machine Learning Platform for AI (PAI). Workspaces allow enterprises and teams to manage computing resources and user permissions in a unified manner. Workspaces provide AI developers with development tools that allow different teams to collaborate throughout the entire workflow of AI development and allow the developers to manage AI computing assets. This topic describes how to create a workspace.
Account and permission requirements
Alibaba Cloud account (recommended): You can use an Alibaba Cloud account to complete all operations without additional authorization.
RAM user: You must grant
AliyunPAIFullAccesspermissions to the RAM user. For more information, see Appendix: AliyunPAIFullAccess. In this case, the RAM user has all the permissions on PAI. Proceed with caution.
Log on to the Machine Learning Platform for AI console.
Go to the Create a new workspace page.
You can go to the Create a new workspace page by using one of the following methods:
On the Overview page, click Create a new workspace.
In the left-side navigation pane, click Workspaces.
On the Workspace list page, click New AI workspace.
In the Basic information step of the Create a new workspace page, set the following parameters and click Next.
The name of the workspace. The name must be 3 to 23 characters in length, and can contain only letters, underscores (_), and digits. The name must start with a letter.Note
If the system prompts that the workspace name already exists but the workspace is not found in the PAI console workspace list, this may be because a workspace with the same name already exists in the DataWorks console. PAI and DataWorks share the same workspaces, therefore you must make sure that each workspace name is unique.
The workspace description. Descriptions are used to identify workspaces.
You can click Add Member to add members to the workspace and assign roles to the members. For more information, see Manage workspace members.
In the left-side navigation pane, clickto view the mappings between roles and permissions and assign appropriate roles to the workspace members.
In the Associated resources step of the Create a new workspace page, set the following parameters and click Create.
Associate Resource Group
Select the computing resources that you want to associate with the workspace. Pay-as-you-go and subscription MaxCompute resources, Deep Learning Containers (DLC) resources, Intelligent Computing Lingjun resources, and Fully Managed Flink resources are supported.
You can chooseto view the resource groups that can be associated with the workspace.Note
If the resource group type that you want to use is not found, contact the resource administrator.
If you select a MaxCompute resource group, you need to specify the MaxCompute project name for the selected resource group. You can also specify whether to use GPU resources. Only pay-as-you-go GPU resources are available.
This section describes the PAI modules that can be used by the resource group.
You can use MaxCompute resource groups, General Training Resources, and Fully Managed Flink resource groups in Machine Learning Designer. As different algorithm components support different computing resources, we recommend that you read up on the introductions of each component in Machine Learning Designer for more information.
In addition, you can submit code and images to the general training resource groups when you create training jobs.
In the Confirm information step of the Create a new workspace page, confirm the information about the workspace and click Enter the workspace.
In the Confirm information step, you can check whether errors exist in the configurations. If errors exist in the configurations, you can return to the previous steps and modify the configurations.
What to do next
After you go to the workspace details page, you can view all the PAI modules in the left-side navigation pane. You can use the modules to manage the full lifecycle of AI development based on your business scenarios. For example, you can refer to the development procedure described in the following figure to use the modules in cloud-native development scenarios.