The recent submission of Google TPU-v3 Pods to the industry wide MLPerf v0.6 training benchmark demonstrates the scalability of a suite of industry relevant ML models. MLPerf defines a suite of models, datasets and rules to follow when benchmarking to ensure results are comparable across hardware, frameworks and companies. Using this suite of models, we discuss the optimizations and techniques including choice of optimizer, spatial partitioning and weight update sharding necessary to scale to 1024 TPU chips. Furthermore, we identify properties of models that make scaling them challenging, such as limited data parallelism and unscaled weights. These optimizations contribute to record performance in transformer, Resnet-50 and SSD in the Google MLPerf-0.6 submission.
@article{arxiv.1909.09756,
title = {Scale MLPerf-0.6 models on Google TPU-v3 Pods},
author = {Sameer Kumar and Victor Bitorff and Dehao Chen and Chiachen Chou and Blake Hechtman and HyoukJoong Lee and Naveen Kumar and Peter Mattson and Shibo Wang and Tao Wang and Yuanzhong Xu and Zongwei Zhou},
journal= {arXiv preprint arXiv:1909.09756},
year = {2019}
}