arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1901.01544cs.LGcs.DCstat.ML
收藏

环AllReduce上重要性加权剪枝的带宽缩减

Bandwidth Reduction using Importance Weighted Pruning on Ring AllReduce

State Key Laboratory of Reliability and Intelligence of Electrical Equipment(可靠性与智能电气设备国家重点实验室);Hebei University of Technology(河北工业大学)

Zehua Cheng, Zhenghua Xu

摘要:

在配备加速器系统的大规模集群上训练大型深度学习模型是必然的。深度梯度压缩可以大大提高带宽利用率,加快训练过程,但难以在环形结构上实现。在本文中,我们发现冗余梯度和梯度老化对训练有负面影响。我们观察到,在不同的时期和不同的步骤,神经网络的重点是更新不同的层和不同的参数。为了节省更多的通信带宽,同时保持环结构的准确性,突破节点增加时的限制,提出了一种基于参数计算梯度与参数值之比大小的大规模集群环全归约梯度重要性度量算法。我们的重要性加权剪枝方法在AlexNet和ImageNet上分别实现了64倍和58.8倍的梯度压缩比。同时,为了保持梯度传播的稀疏性,我们在每个节点上随机广播重要梯度的索引。而其余节点准备好索引梯度并执行all-reduce更新。这将加快模型的收敛速度并保持训练精度。

英文摘要:

It is inevitable to train large deep learning models on a large-scale cluster equipped with accelerators system. Deep gradient compression would highly increase the bandwidth utilization and speed up the training process but hard to implement on ring structure. In this paper, we find that redundant gradient and gradient staleness has negative effect on training. We have observed that in different epoch and different steps, the neural networks focus on updating different layers and different parameters. In order to save more communication bandwidth and preserve the accuracy on ring structure, which break the restrict as the node increase, we propose a new algorithm to measure the importance of gradients on large-scale cluster implementing ring all-reduce based on the size of the ratio of parameter calculation gradient to parameter value. Our importance weighted pruning approach achieved 64X and 58.8X of gradient compression ratio on AlexNet and ResNet50 on ImageNet. Meanwhile, in order to maintain the sparseness of the gradient propagation, we randomly broadcast the index of important gradients on each node. While the remaining nodes are ready for the index gradient and perform all-reduce update. This would speed up the convergence of the model and preserve 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。

↑