SiP Architecture For Accelerating Collective Communication in Distributed Deep Learning
暂无分享,去创建一个
We present a silicon photonic architecture for accelerating collective communications in distributed deep learning. We demonstrate a 22% job completion time improvement in a small-scale testbed and 1.4 to 5.9× improvement in large-scale simulations.
[1] S. Yoo,et al. Silicon Photonic Flex-LIONS for Reconfigurable Multi-GPU Systems , 2021, Journal of Lightwave Technology.
[2] Haitao Wu,et al. BCube: a high performance, server-centric network architecture for modular data centers , 2009, SIGCOMM '09.