Better ML for Everyone with ML Commons

ML Commons aims to build better machine learning for everyone through three pillars. We create benchmarks like the MLPerf suite that measure speed for ML training and inference from microwatts to megawatts. We build large, open, and public datasets such as The People’s Speech (30K hours of labeled speech), Multilingual Spoken Words Corpus (keyword spotting for 23M utterances in 50 languages) to empower the community to innovate. Last, we build best practices that reduce friction like MLCube, which provides a common container interface for ML models making them easier to share and use.