Participating in upcoming optimization challenges to co-design Pareto-efficient AI/ML systems using the MLCommons automation platform

1 view

Skip to first unread message

Grigori Fursin

unread,

Apr 25, 2023, 7:41:34 AM4/25/23

to Collective Knowledge, Arjun Suresh

Dear all,

Following our presentation at the MLCommons community meeting about our new MLCommons CK-MLPerf platform to co-design Pareto-efficient AI/ML systems via reproducible optimization tournaments, we have received many questions about how to participate in this project. Thank you very much for your interest - we have aggregated all your feedback into one short questionnaire available at the following link:

https://forms.gle/VWw4FxRkpgPsVith9

Please fill it in to let us know how you would like to contribute.

Note that this is a heavily evolving project: we suggest you join the MLCommons taskforce on automation and reproducibility via our public Discord server and weekly conf-calls. Our ultimate goal is to help anyone automatically co-design Pareto-efficient AI/ML Systems and end-to-end AI/ML applications in a unified, automated, collaborate and reproducible way based on their requirements and constraints (usage cost vs performance vs power consumption vs accuracy etc) from any suitable model, data, software and hardware from any vendor.

Looking forward to collaborating with you,

Grigori Fursin and Ajrun Suresh

————————————————————————————————

Grigori Fursin

* Co-founder of the MLCommons taskforce on automation and reproducibility

* President of the cTuning foundation

* Founding member of the ACM SIG on reproducible research