arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2009.02701cs.LGstat.ML
收藏

HPSGD:具有陈旧梯度的分层并行SGD

HPSGD: Hierarchical Parallel SGD With Stale Gradients Featuring

Sichuan University(四川大学)

Yuhao Zhou, Qing Ye, Hailun Zhang, Jiancheng Lv

摘要:

虽然分布式训练显著加快了深度神经网络(DNN)的训练过程,但由于工作者之间的数据同步耗时,集群的利用率相对较低。为了缓解这个问题,提出了一种新的分层并行SGD(HPSGD)策略,该策略基于数据同步阶段可以与本地训练阶段(即,前馈和反向传播)。此外,改进的模型更新方法被单元化以补救引入的陈旧梯度问题,其将更新提交到副本(即,具有与全局模型相同的参数的临时模型),然后将平均变化合并到全局模型。大量的实验表明,所提出的HPSGD方法大大提高了分布式DNN训练,减少了陈旧梯度的干扰,并在给定的固定壁时间内实现了更好的准确性。

英文摘要:

While distributed training significantly speeds up the training process of the deep neural network (DNN), the utilization of the cluster is relatively low due to the time-consuming data synchronizing between workers. To alleviate this problem, a novel Hierarchical Parallel SGD (HPSGD) strategy is proposed based on the observation that the data synchronization phase can be paralleled with the local training phase (i.e., Feed-forward and back-propagation). Furthermore, an improved model updating method is unitized to remedy the introduced stale gradients problem, which commits updates to the replica (i.e., a temporary model that has the same parameters as the global model) and then merges the average changes to the global model. Extensive experiments are conducted to demonstrate that the proposed HPSGD approach substantially boosts the distributed DNN training, reduces the disturbance of the stale gradients and achieves better accuracy in given fixed wall-time.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑