A model is one of the core objects in MaxCompute. MaxCompute supports various model types, including public, imported, and remote models. It provides unified management for models and model versions, helping you seamlessly integrate AI capabilities into your business analysis workflows. This topic describes the concepts, benefits, and types of MaxCompute models, and explains how to manage and use them.
Model overview
Key concepts
model: A deployment object registered in MaxCompute for prediction or generation tasks. It seamlessly integrates AI capabilities, such as large language models (LLMs) and machine learning models, into the platform where your data resides.
model version: An independent and uniquely identifiable sub-object of a model. You can create and manage multiple iterative versions under the same model name. This simplifies gradual rollouts, rapid rollbacks, and side-by-side performance comparisons when you call models by using an AI Function.
Benefits
Unified management: MaxCompute provides multiple model types, which, like data, support permission control and versioning to meet enterprise security and compliance requirements.
Multi-engine support: You can call models from various ecosystems, such as SQL and Python (MaxFrame). This unified architecture not only allows data analysts to use familiar SQL to call powerful AI models but also enables data scientists to use the distributed Python computing power of MaxFrame to improve data preprocessing efficiency and quality.
Simplified O&M: You can perform AI inference without exporting data. This eliminates the security risks, costs, and latency of data movement.
Model types
MaxCompute provides the following types of models:
Model type | Description | Tutorial |
Public model |
| Use a MaxCompute public model for sentiment analysis of online reviews |
Remote model | Connect to models deployed on PAI-EAS. Register them as MaxCompute remote models by providing the PAI-EAS endpoint and access token. You can then call the models using an AI Function. | Use a MaxCompute remote model to automatically generate e-commerce product selection descriptions |
Internally trained model | You can use MaxCompute MaxFrame to train traditional machine learning models. Then, run | |
Imported model | When built-in public models are insufficient, you can import externally fine-tuned models for better performance tailored to your business needs. Import custom, externally trained models by specifying their OSS path. MaxCompute can then use them for inference. | Being rolled out |
Use the MaxCompute AI Function to call built-in public models or other models you have created.
Public models and applicable quotas
Public model name | Applicable quota |
|
|
|
|
|
|
Model management
Before managing models, ensure your account has the required permissions to manage model objects.
You can manage model objects in the following ways:
Method
Description
Create, view, modify, and delete models using SQL statements.
Create models using the MaxFrame Python language. This method currently only supports model creation.
Manage models in the console
Manage models through the console's graphical user interface (GUI). In supported regions, you can view your models.
Follow these steps:
Log in to the MaxCompute console and select a region in the upper-left corner.
In the left-side navigation pane, choose .
On the Projects page, click Manage in the Actions column for the target project.
On the Project Settings page, click the Models tab.
You can view public models and their versions in the
BIGDATA_PUBLIC_MODELSETpublic project, or view your own models and their versions.
NoteThe console-based model management feature is currently available only in the China (Beijing), China (Hangzhou), China (Shanghai), and China (Shenzhen) regions. Support for other regions is rolling out.