arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2102.03112cs.LG
收藏

DeepReduce:一个用于分布式深度学习的稀疏张量通信框架

DeepReduce: A Sparse-tensor Communication Framework for Distributed Deep Learning

Columbia University(哥伦比亚大学);KAUST(阿卜杜拉国王科技大学);University of Central Florida(中佛罗里达大学)

Kelly Kostopoulou, Hang Xu, Aritra Dutta, Xin Li, Alexandros Ntoulas, Panos Kalnis

摘要:

稀疏张量经常出现在分布式深度学习中,要么是深度神经网络梯度的直接伪影,要么是显式稀疏化过程的结果。现有的通信原语对深度学习的特性是不可知的;因此,它们会带来不必要的通信开销。本文介绍了DeepReduce,这是一个用于稀疏张量压缩通信的通用框架,专为分布式深度学习量身定制。DeepReduce将稀疏张量分解为两个集合,值和索引,并允许对这些集合进行独立和组合压缩。我们支持各种常见的压缩器,例如用于值的Deflate,或用于索引的行程长度编码。我们还提出了两种新的压缩方案,实现卓越的结果:基于曲线拟合的值和基于布隆过滤器的指数。DeepReduce与现有的梯度稀疏器正交,可以与它们一起应用,对最终用户透明,以显着降低通信开销。作为概念证明,我们在Tensorflow和PyTorch上实现了我们的方法。我们对大型真实模型的实验表明,DeepReduce传输的数据更少,比现有方法的计算开销更低,而不会影响训练精度。

英文摘要:

Sparse tensors appear frequently in distributed deep learning, either as a direct artifact of the deep neural network's gradients, or as a result of an explicit sparsification process. Existing communication primitives are agnostic to the peculiarities of deep learning; consequently, they impose unnecessary communication overhead. This paper introduces DeepReduce, a versatile framework for the compressed communication of sparse tensors, tailored for distributed deep learning. DeepReduce decomposes sparse tensors in two sets, values and indices, and allows both independent and combined compression of these sets. We support a variety of common compressors, such as Deflate for values, or run-length encoding for indices. We also propose two novel compression schemes that achieve superior results: curve fitting-based for values and bloom filter-based for indices. DeepReduce is orthogonal to existing gradient sparsifiers and can be applied in conjunction with them, transparently to the end-user, to significantly lower the communication overhead. As proof of concept, we implement our approach on Tensorflow and PyTorch. Our experiments with large real models demonstrate that DeepReduce transmits fewer data and imposes lower computational overhead than existing methods, without affecting the training accuracy.

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。

↑