MLPerf Training Benchmark
This provides a standardized benchmark for evaluating ML software and hardware, addressing a critical need for fair and effective performance comparison in the industry.
The paper tackles the lack of industry-standard performance benchmarks for machine learning training by presenting MLPerf, a benchmark designed to address unique challenges like optimization trade-offs, stochasticity, and system diversity, with analysis showing its efficacy in driving performance and scalability improvements across multiple vendors.
Machine learning (ML) needs industry-standard performance benchmarks to support design and competitive evaluation of the many emerging software and hardware solutions for ML. But ML training presents three unique benchmarking challenges absent from other domains: optimizations that improve training throughput can increase the time to solution, training is stochastic and time to solution exhibits high variance, and software and hardware systems are so diverse that fair benchmarking with the same binary, code, and even hyperparameters is difficult. We therefore present MLPerf, an ML benchmark that overcomes these challenges. Our analysis quantitatively evaluates MLPerf's efficacy at driving performance and scalability improvements across two rounds of results from multiple vendors.