Improving Efficiency in Large-Scale Decentralized Distributed Training

February 04, 2020 · Declared Dead · 🏛 IEEE International Conference on Acoustics, Speech, and Signal Processing

"No code URL or promise found in abstract"

Evidence collected by the PWNC Scanner

Authors Wei Zhang, Xiaodong Cui, Abdullah Kayi, Mingrui Liu, Ulrich Finkler, Brian Kingsbury, George Saon, Youssef Mroueh, Alper Buyuktosunoglu, Payel Das, David Kung, Michael Picheny arXiv ID 2002.01119 Category cs.LG: Machine Learning Cross-listed cs.DC, stat.ML Citations 16 Venue IEEE International Conference on Acoustics, Speech, and Signal Processing Last Checked 4 months ago

Abstract

Decentralized Parallel SGD (D-PSGD) and its asynchronous variant Asynchronous Parallel SGD (AD-PSGD) is a family of distributed learning algorithms that have been demonstrated to perform well for large-scale deep learning tasks. One drawback of (A)D-PSGD is that the spectral gap of the mixing matrix decreases when the number of learners in the system increases, which hampers convergence. In this paper, we investigate techniques to accelerate (A)D-PSGD based training by improving the spectral gap while minimizing the communication cost. We demonstrate the effectiveness of our proposed techniques by running experiments on the 2000-hour Switchboard speech recognition task and the ImageNet computer vision task. On an IBM P9 supercomputer, our system is able to train an LSTM acoustic model in 2.28 hours with 7.5% WER on the Hub5-2000 Switchboard (SWB) test set and 13.3% WER on the CallHome (CH) test set using 64 V100 GPUs and in 1.98 hours with 7.7% WER on SWB and 13.3% WER on CH using 128 V100 GPUs, the fastest training time reported to date.