This repository contains the automations and scripts used to run MLPerf benchmarks, primarily focusing on MLPerf inference benchmarks. The automations used here are largely based on and extended from the Collective Mind script automations.
CM (Collective Mind) is a Python package with a CLI and API designed to create and manage automations. Two key automations developed using CM are Script and Cache, which streamline ML workflows, including managing Docker runs.
The CM Python package was developed by Grigori Fursin. The Script and Cache automations are part of the cm4mlops repository, created by Grigori Fursin and Arjun Suresh and sponsored by OctoML, cKnowledge, cTuning and MLCommons.
The CM scripts, also housed in the cm4mlops repository, are created and maintained by Arjun Suresh and Anandhu Sooraj, and Grigori Fursin with the help of the MLCommons community.
CM CLI: https://docs.mlcommons.org/ck/specs/cm-cli/
Documentation site for MLPerf Inference: https://docs.mlcommons.org/inference/