View and use built-in images - MaxCompute - Alibaba Cloud Documentation Center

MaxCompute provides built-in images for data analytics, scientific computing, and machine learning, such as Pandas, Numpy, Scikit-learn, and Xgboost. You can reference these images directly to simplify your development process. This topic describes how to view and use built-in images.

View built-in images

Log on to the MaxCompute console and select a region in the top-left corner.
In the navigation pane on the left, choose Manage Configurations > Images.
On the Images page, click the Built-in Images tab.
You can view the list of built-in MaxCompute images and their version information.

Built-in image details

Category	Image name	Built-in third-party package and version	Programming language and version	Description
Base image	common	numpy==1.21.6 pandas==1.3.5	Python 3.7	The base image. It includes basic third-party packages such as Pandas and NumPy.
Base image	common	numpy==1.26.4 pandas==2.2.2	Python 3.11
Base image	ubuntu	ubuntu==20.04	Python 3.7 Python 3.11	A base operating system (OS) image. It supports seamless use in user-defined function (UDF) development and MaxFrame framework integration. This improves cross-platform code portability and execution performance.
Scientific computing	scipy	scipy==1.7.3	Python 3.7	A scientific computing library that provides advanced scientific computing features, including statistical analysis and linear algebra.
Scientific computing	scipy	scipy==1.13.0	Python 3.11
Statistical modeling	statsmodels	statsmodels==0.13.5	Python 3.7	Statistical modeling
Statistical modeling	statsmodels	statsmodels==0.14.1	Python 3.11	A library for statistical modeling and econometrics.
Machine learning	sklearn	scikit-learn==1.0.2	Python 3.7	Provides machine learning algorithms such as classification, regression, and clustering.
	sklearn	scikit-learn==1.4.2	Python 3.11
	xgboost	xgboost==1.6.2	Python 3.7	A distributed gradient boosting library.
	xgboost	xgboost==2.0.3	Python 3.11	A distributed gradient boosting library.
	pytorch	torch==1.13.1	Python 3.7	Natural Language Processing.
	pytorch	torch==2.3.0	Python 3.11	Natural Language Processing.
	tensorflow	tensorflow==2.11.0	Python 3.7	Used for programming various machine learning algorithms.
	tensorflow	tensorflow==2.16.1	Python 3.11	Used for programming various machine learning algorithms.

All built-in images include development packages from the base image, such as numpy and pandas. They also include common basic development packages, such as cloudpickle 2.2.1, pickle 5.0.12, requests 2.31.0, and setuptools 68.0.0.

Use built-in images

You can use built-in images in MaxCompute SQL user-defined function (UDF), PyODPS, or MaxFrame development.

Important

To prevent conflicts, specify only one image for each development job.

When you call a UDF, specify the dependent image and Python version at the SQL session level using a flag. Use the following commands:

-- If you use Python 3.11, set odps.sql.python.version=cp311;
set odps.sql.python.version=cp37;
set odps.session.image = <image_name>;

In PyODPS development, specify an existing image using the image parameter of the execute or persist method. For an example, see Use images in PyODPS development. Use the following command:
```
image='<image_name>'
```
Note
To reference an image for PyODPS development, upgrade PyODPS to V0.11.5 or a later version.
In MaxFrame development, specify an existing image for the current job. The relevant parameters are as follows:
```
config.options.sql.settings = {
    "odps.session.image": "<image_name>"
}
```