Python Function Library Reference

Artikel
07/15/2019

This section contains the Python reference documentation for three proprietary packages from Microsoft used for data science and machine learning on premises and at scale.

You can use these libraries and functions in combination with other open source or third-party packages, but to use the proprietary packages, your Python code must run against a service or on a computer that provides the interpreters.

Library details	Description
Supported platforms	Machine Learning Server 9.2.1, 9.3 and 9.4 SQL Server 2017 (Windows only)
Built on:	Anaconda 4.2 distribution of Python 3.5 (included when you add Python support during installation).

Python modules

Module	Version	Description
azureml-model-management-sdk	1.0.1	Classes and functions to authenticate, deploy, manage, and consume analytic web services in Python.
microsoftml	9.4	A collection of Python functions used for machine learning operations, including training and transformations, scoring, text and image analysis, and feature extraction for deriving values from existing data.
revoscalepy	9.4	Data access, manipulation and transformations, visualization, and statistical analysis. The revoscalepy functions support a broad spectrum of statistical and analytical tasks that operate at scale, bringing analytical operations to local data or remote on a Spark cluster or data residing in SQL Server.

Note

Developers who are familiar with Microsoft R packages might notice similarities in the functions provided in revoscalepy and microsoftml. Conceptually, revoscalepy and microsftml are the Python equivalents of the RevoScaleR R package and the MicrosoftML R package, respectively.

How to get packages

You can get the packages when you install Machine Learning Server, or SQL Server 2017, and choose the option for Python support. In addition to Python packages, Machine Learning Server setup and SQL Server setup both install the Python interpreters and base modules required to run any script or code that calls functions from proprietary package.

For SQL Server, packages are installed by default in the \Program files\Microsoft SQL Server*instance name*\PYTHON_SERVICES

Ships in:

How to list modules and versions

To get the version of a Python module installed on your computer, start Python and execute the following commands:

Double-click Python.exe in \Program Files\Microsoft\ML Server\PYTHON_SERVER.
Open interactive help: help()
Type the name of a module at the help prompt: help> revoscalepy. Help returns the name, package contents, version, and file location.
List all installed modules: modules
Import a module: import revoscalepy

How to list functions and get function help

To view the embedded help for each class, use the help() command, specifying the base class of the object of interest.

Double-click Python.exe in \Program Files\Microsoft\ML Server\PYTHON_SERVER.
Open interactive help: help()
Type the fully-qualified class name within the brackets.
- For azureml-azureml-model-management-sdk, include the class in the path. For example, for MLServer help, type help(azureml.deploy.server.MLServer).
- For revoscalepy, type help(revoscalepy) to get package contents, and then include one of the packages on the next iteration. For example, help(revoscalepy.computecontext) returns all the functions related to compute context.

Note to R Users: Python naming conventions

revoscalepy, microsoftml, and azureml-model-management-sdk correspond to the Microsoft R packages, RevoScaleR, MicrosoftML, and mrsdeploy. If you have a background in these libraries, you might notice similarities in function names and operations, with Python versions adhering to the naming conventions of that language:

lowercase package names (microsoftml contrasted with MicrosoftML) and most function names
underscore in function names (rx_import in revoscalepy contrasted with rxImport in RevoScaleR)

Next steps

First, read the introduction to each package to learn about common use case scenarios:

Next, follow these tutorials for hands-on experience:

Freigeben über