Skip to content

A collection of reusable and cross-platform automation recipes (CM scripts) with a human-friendly interface and minimal dependencies to make it easier to build, run, benchmark and optimize AI, ML and other applications and systems across diverse and continuously changing models, data sets, software and hardware (cloud/edge)

License

Notifications You must be signed in to change notification settings

GATEOverflow/cm4mlops

 
 

Repository files navigation

Unified and cross-platform CM interface for DevOps, MLOps and MLPerf

License Python Version Powered by CM. Downloads

CM script automation features test MLPerf inference bert (deepsparse, tf, onnxruntime, pytorch) MLPerf inference MLCommons C++ ResNet50 MLPerf inference ABTF POC Test Test Compilation of QAIC Compute SDK (build LLVM from src) Test QAIC Software kit Compilation

Collective Mind (CM)

Collective Mind (CM) is a Python package with a CLI and API designed for creating and managing automations. Two key automations developed using CM are Script and Cache, which streamline machine learning (ML) workflows, including managing Docker runs. Both Script and Cache automations are part of the cm4mlops repository.

The CM scripts, also housed in the cm4mlops repository, consist of hundreds of modular Python-wrapped scripts accompanied by yaml metadata, enabling the creation of robust and flexible ML workflows.

The mlperf-branch of the cm4mlops repository is dedicated to developments specific to MLPerf Inference. Please submit any pull requests (PRs) to this branch. For more information about using CM for MLPerf Inference, refer to the MLPerf Inference Documentation.

News

License

Apache 2.0

CM concepts

Check our ACM REP'23 keynote.

Authors

Grigori Fursin and Arjun Suresh

Major script developers

Arjun Suresh, Anandhu S, Grigori Fursin

Funding

We thank cKnowledge.org, cTuning foundation and MLCommons for sponsoring this project!

Acknowledgments

We thank all volunteers, collaborators and contributors for their support, fruitful discussions, and useful feedback!

About

A collection of reusable and cross-platform automation recipes (CM scripts) with a human-friendly interface and minimal dependencies to make it easier to build, run, benchmark and optimize AI, ML and other applications and systems across diverse and continuously changing models, data sets, software and hardware (cloud/edge)

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 75.0%
  • Shell 12.2%
  • C++ 7.1%
  • C 2.9%
  • Batchfile 2.0%
  • Dockerfile 0.5%
  • Other 0.3%