Scale MLPerf-0.6 models on Google TPU-v3 Pods
This work addresses the problem of efficiently scaling industry-relevant ML models for hardware benchmarking, though it is incremental as it applies known techniques to a new hardware setup.
The paper tackled scaling MLPerf-0.6 models on Google TPU-v3 Pods, achieving record performance in transformer, ResNet-50, and SSD benchmarks by applying optimizations like spatial partitioning and weight update sharding to scale to 1024 TPU chips.
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.