Welcome to the MLPerf Automations and Scripts repository! This repository is your go-to resource for tools, automations, and scripts designed to streamline the execution of MLPerf benchmarks—with a strong emphasis on MLPerf Inference benchmarks.
Starting January 2025, MLPerf automation scripts are built on the powerful MLCFlow automation interface. This modern interface replaces the earlier Collective Mind (CM), offering a more robust and efficient framework for benchmarking workflows.
- Automated Benchmarking – Simplifies running MLPerf Inference benchmarks with minimal manual intervention.
- Modular and Extensible – Easily extend the scripts to support additional benchmarks and configurations.
- Seamless Integration – Compatible with Docker, cloud environments, and local machines.
- MLCFlow (MLC) Integration – Utilizes the MLC framework to enhance reproducibility and automation.
Building upon the robust foundation of its predecessor, the Collective Mind (CM) framework, MLCFlow elevates machine learning workflows by simplifying complex tasks such as Docker container management and caching. Written in Python, the mlcflow package offers a versatile interface, supporting both a user-friendly command-line interface (CLI) and a flexible API for effortless automation script management.
At its core, MLCFlow relies on a single powerful automation, the Script, which is extended by two actions: CacheAction and DockerAction. Together, these components provide streamlined functionality to optimize and enhance your ML workflow automation experience.
We welcome contributions from the community! To contribute:
- Submit pull requests (PRs) to the
devbranch. - Review our CONTRIBUTORS.md for guidelines and best practices.
- Explore more about MLPerf Inference automation in the official MLPerf Inference Documentation.
Your contributions help drive the project forward!
Connect with us on the MLCommons Benchmark Infra Discord channel to engage in discussions about MLCFlow and MLPerf Automations. We’d love to hear your thoughts, questions, and ideas!
Keep track of the latest development progress and tasks on our MLPerf Automations Development Board.
Stay tuned for exciting updates and announcements!
This project is licensed under the Apache 2.0 License.
This project is made possible through the generous support of:
We appreciate their contributions and sponsorship!
Thank you for your interest and support in MLPerf Automations and Scripts!