PAI covers the end-to-end AI development lifecycle, from data preparation and model training to model deployment. Read this topic to quickly understand the PAI components and find the right path to get started.
1. Activate PAI
Log in to the PAI console. In the upper-left corner, select a region and click Activate. After you activate the service, PAI automatically creates a default workspace.
We recommend activating PAI with an Alibaba Cloud account to prevent activation failures due to a missing AliyunPAIFullAccess policy.2. Quick start
New users can start with a single component. Select a suitable component based on the descriptions in the following table.
After you select a component, read its quick start guide, which provides simple examples to help you quickly get started.
Read the user guide for the component to learn more about its features and best practices.
Click a link in the **Component** column to go to the user guide for that component.
Component | Description | Quick start |
Integrates DLC and EAS to help you efficiently train and deploy open-source large models without code. | ||
Deploy trained models as online inference services with minimal configuration. | ||
Provides a cloud-based IDE for AI development (development instance). Developers familiar with Notebook or VSCode can quickly start model development. | ||
Allows you to quickly create distributed or single-node training tasks. This eliminates the need to manually purchase machines and configure runtime environments, and matches the experience of running training scripts locally. | ||
Provides more than 140 built-in algorithm components for building models visually with a low-code, drag-and-drop approach. |
3. Get help
Read the FAQ
The FAQ lists common issues that developers encounter when using PAI and provides their solutions. If you encounter issues when using components such as DSW and EAS, refer to the relevant FAQ:
Consult the AI Assistant
The AI Assistant in the lower-right corner of the official website provides real-time, accurate answers to your questions about cloud products, offering quick access to technical support and documentation.
4. Typical AI development workflows
PAI supports the end-to-end AI development process, from data preparation and model training to model deployment. The following sections describe two typical AI development workflows.
Cloud-native AI development
Step | Description | Related documentation |
① | PAI's dataset management feature lets you centrally manage local, cloud, and public datasets as data sources for model training. | |
② | DSW provides a cloud-based IDE for AI development. Developers who are familiar with Notebook or VSCode can quickly start model development. | |
③ | Images provide the runtime environment for code execution. PAI's image management feature lets you centrally manage official public images and custom images. | |
④ | After you complete model code development and testing in DSW, you can use DLC to run training tasks for higher efficiency and cost savings. | |
⑤ | PAI supports mounting a file system (NAS or OSS) and Git repositories, which makes it easy to specify data and code when you submit a task. | |
⑥ | The model management feature centrally manages trained models, allowing you to use them directly for EAS model deployment. | |
⑦ | After model training is complete, you can use EAS to quickly deploy the model as an online service. |
AI and big data development
Step | Description | Related documentation |
① | If you use MaxCompute to store data, preprocess the data in DataWorks first, and then use the MaxCompute table in PAI as a data source for training. | |
② | Designer provides more than 140 built-in algorithm components for building models with low code in a drag-and-drop visual interface. | |
③ | DataWorks lets you configure and run periodic scheduled tasks. | |
④ | The task management service records execution information for Designer experiments and custom tasks, simplifying the comparison and analysis of the results of different tasks. | |
⑤ | The model management feature centrally manages trained models, allowing you to use them directly for EAS model deployment. | |
⑥ | After model training is complete, you can use EAS to quickly deploy the model as an online service. |
5. FAQ
Q: Why is the Activate Now button disabled?
Use one of the following methods:
Use your Alibaba Cloud account to complete the activation.
If you are using a RAM user, you must attach the AliyunPAIFullAccess system policy to the RAM user. Note: This policy grants extensive permissions. The Alibaba Cloud account administrator must evaluate the security risks before granting this policy.
Q: How to fix the error: Create order error: message is Your account balance is less than 0. Please top up your account and try to purchase again. productRequestId is ***?
Go to User Center, view your bills and pay any outstanding fees, and then try to activate PAI again.
Q: How to fix the error: Create order error: message is Failed to verify the order before placing productRequestId is 8F5?
This error indicates insufficient permissions. Use an Alibaba Cloud account or a RAM user with the AliyunPAIFullAccess policy to perform the operation.