# A Highly Efficient Distributed Deep Learning System For Automatic Speech Recognition

## TL;DR

This paper introduces a highly efficient distributed deep learning system for automatic speech recognition that leverages a novel ADPSGD algorithm, enabling faster training with larger batch sizes and achieving state-of-the-art results in a short time.

## Contribution

The work demonstrates that ADPSGD can handle much larger batch sizes than SSGD, and proposes a hierarchical system that significantly accelerates ASR training at scale.

## Key findings

- ADPSGD converges with 3X larger batch sizes than SSGD.
- The hierarchical system trains SWB-2000 in 5.2 hours with high accuracy.
- Achieved the fastest reported training time for SWB-2000 with competitive WERs.

## Abstract

Modern Automatic Speech Recognition (ASR) systems rely on distributed deep learning to for quick training completion. To enable efficient distributed training, it is imperative that the training algorithms can converge with a large mini-batch size. In this work, we discovered that Asynchronous Decentralized Parallel Stochastic Gradient Descent (ADPSGD) can work with much larger batch size than commonly used Synchronous SGD (SSGD) algorithm. On commonly used public SWB-300 and SWB-2000 ASR datasets, ADPSGD can converge with a batch size 3X as large as the one used in SSGD, thus enable training at a much larger scale. Further, we proposed a Hierarchical-ADPSGD (H-ADPSGD) system in which learners on the same computing node construct a super learner via a fast allreduce implementation, and super learners deploy ADPSGD algorithm among themselves. On a 64 Nvidia V100 GPU cluster connected via a 100Gb/s Ethernet network, our system is able to train SWB-2000 to reach a 7.6% WER on the Hub5-2000 Switchboard (SWB) test-set and a 13.2% WER on the Call-home (CH) test-set in 5.2 hours. To the best of our knowledge, this is the fastest ASR training system that attains this level of model accuracy for SWB-2000 task to be ever reported in the literature.