arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2107.01739cs.LGcs.DC
收藏

KAISA:深度神经网络的自适应二阶优化器框架

KAISA: An Adaptive Second-Order Optimizer Framework for Deep Neural Networks

University of Chicago(芝加哥大学);University of Texas at Austin(得克萨斯大学奥斯汀分校);Texas Advanced Computing Center(得克萨斯高级计算中心);University of Wisconsin, Madison(威斯康星大学麦迪逊分校);Argonne National Laboratory(阿贡国家实验室)

J. Gregory Pauloski, Qi Huang, Lei Huang, Shivaram Venkataraman, Kyle Chard, Ian Foster, Zhao Zhang

摘要:

最近,Kronecker因子近似曲率(K-FAC)在深度神经网络(DNN)训练中的收敛速度比随机梯度下降(SGD)更快;然而,K-FAC较大的内存占用阻碍了其对大型模型的适用性。我们提出了KAISA,一个K-FAC启用,适应性强,改进和可扩展的二阶优化器框架,它适应特定模型和硬件的内存占用,通信和计算,以提高性能并增加可扩展性。我们量化了内存和通信成本之间的权衡,并在大型模型上评估KAISA,包括ResNet-50,Mask R-CNN,U-Net和BERT,最多128个NVIDIA A100 GPU。与原始优化器相比,KAISA在相同全局批量大小的应用程序中收敛速度快18.1-36.3%。在固定的内存预算下,KAISA在ResNet-50和BERT-Large中的收敛速度分别提高了32.5%和41.6%。KAISA可以平衡内存和通信,以实现等于或优于基线优化器的扩展效率。KAISA是开源的,可在https://github.com/gpauloski/kfac_pytorch上获得。

英文摘要:

Kronecker-factored Approximate Curvature (K-FAC) has recently been shown to converge faster in deep neural network (DNN) training than stochastic gradient descent (SGD); however, K-FAC's larger memory footprint hinders its applicability to large models. We present KAISA, a K-FAC-enabled, Adaptable, Improved, and ScAlable second-order optimizer framework that adapts the memory footprint, communication, and computation given specific models and hardware to improve performance and increase scalability. We quantify the tradeoffs between memory and communication cost and evaluate KAISA on large models, including ResNet-50, Mask R-CNN, U-Net, and BERT, on up to 128 NVIDIA A100 GPUs. Compared to the original optimizers, KAISA converges 18.1-36.3% faster across applications with the same global batch size. Under a fixed memory budget, KAISA converges 32.5% and 41.6% faster in ResNet-50 and BERT-Large, respectively. KAISA can balance memory and communication to achieve scaling efficiency equal to or better than the baseline optimizers. KAISA is open source and available at https://github.com/gpauloski/kfac_pytorch.

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。

↑