Inference Working Group
- Training Working Group
- Inference Working Group
- Datasets Working Group
- Best Practices Working Group
- Research Working Group
Create a set of fair and representative inference benchmarks.
Machine-learning (ML) hardware and software system demand is burgeoning. Driven by ML applications, the number of different ML inference systems has exploded. Over 100 organizations are building ML inference chips, and the systems that incorporate existing models span at least three orders of magnitude in power consumption and five orders of magnitude in performance; they range from embedded devices to data-center solutions. Fueling the hardware are a dozen or more software frameworks and libraries. The myriad combinations of ML hardware and ML software make assessing ML-system performance in an architecture-neutral, representative, and reproducible manner challenging. There is a clear need for industry-wide standard ML benchmarking and evaluation criteria. MLPerf™ Inference answers that call.
- Inference benchmark rules and definitions
- Inference benchmark reference software
- Inference benchmark submission rules
- Inference benchmark roadmap
- Publish inference benchmark results every ~6 months
Weekly on Tuesday from 8:30-10:30AM Pacific.
Working Group Resources
Working Group Chair Emails
Ramesh Chukka (firstname.lastname@example.org)
Tom Jablin (email@example.com)
Working Group Chair Bios
Ramesh Chukka is a Deep Learning Manager at Intel with focus on performance analysis and benchmarking. He has 14+ years of experience leading benchmark development and working with industry benchmark consortiums. Ramesh received M.Tech from IIT Madras and B.E from Andhra University, India.
Tom Jablin is a Staff Software Engineer working on the ML Inference Performance team at Google. He is interested in developing metrics that accurately reflect the experiences of real inference customers and supporting diverse and innovative computer architectures.