Power Working Group
Mission
Create power measurement techniques for various MLPerf benchmarks that enable reporting and comparing energy consumption, performance and power of benchmarks run on submission systems.
Purpose
Power consumption and energy efficiency are critical challenges for deploying and operating machine learning systems across the spectrum, from battery-powered smartphones to the world’s largest data centers. The Power working group will create tools to measure power for machine learning systems to evaluate efficiency and guide system optimization and design trade-offs.
Deliverables
- Power measurement techniques built on industry-standard tools
- List of approved power analyzers
- Power result metrics and format
- Initial deliverable is integration with MLPerf Inference v1.0 for wall-powered systems
- Roadmap for battery-powered system and MLPerf Training
Join
Meeting Schedule
Tuesday November 12, 2024
Weekly – 15:05 – 16:00 Pacific Time
Related Blog
-
New MLPerf Inference v4.1 Benchmark Results Highlight Rapid Hardware and Software Innovations in Generative AI Systems
New mixture of experts benchmark tracks emerging architectures for AI models
-
New MLPerf Training Benchmark Results Highlight Hardware and Software Innovations in AI Systems
Two new benchmarks added – highlighting language model fine-tuning and classification for graph data
-
MLPerf Inference v1.0 Results with First Power Measurements
The latest benchmark includes 1,994 performance and 862 power efficiency results for leading ML inference systems
How to Join and Access Power Working Group Resources
The Power working group is limited exclusively to MLCommons members and affiliates. If you are not already a member or affiliate or part of a member or affiliate company, you can learn more about MLCommons membership here.
- To sign up for the group mailing list, receive the meeting invite, and access shared documents and meeting minutes:
- Fill out our subscription form and indicate that you’d like to join the Power Working Group.
- Associate a Google account with your organizational email address.
- Once your request to join the Power Working Group is approved, you’ll be able to access the Power folder in the Members Google Drive.
- To engage in working group discussions, join the working group’s channels on the MLCommons Discord server.
- To access the GitHub repository (public):
- If you want to contribute code, please submit your GitHub ID to our subscription form.
- Visit the GitHub repository.
Power Working Group Chairs
To contact all Power working group chairs email [email protected].
Arun Tejus Raghunath Rajan
Tejus is a Technical Lead at Intel influencing Intel Products and IPs in the AI and HPC space. He has been at Intel since Summer 2011 and has experience in power modeling, post silicon power analysis and setting requirements for future Intel products. He has had experience in the Small Form Factor and wearables segments as well during this time. He has also served as the President of an Intel Employee Resource Groups helping drive mentorship and D&I efforts within the company. Prior to Intel, he was a Product Engineer at a university startup. Tejus has done his Master’s from the University of Utah in Salt Lake City and Bachelor’s from SRM University in India. Outside of work, he continues his passion for numbers and modeling by participating in soccer fantasy leagues and is an avid soccer fan.
Anirban Ghosh
Anirban Ghosh is a Senior Deep Learning Architect at NVIDIA focusing on HW and SW optimizations for High Performance AI Computing on GPUs and datacenter systems. Previously, he worked as a Field Application Engineer at Texas Instruments. He holds a master’s degree in Electrical & Computer Engineering from Carnegie Mellon University, and a bachelor’s degree from the National Institute of Technology Karnataka in India.