GradPIM: A Practical Processing-in-DRAM Architecture for Gradient Descent
arXiv:2102.07511
Abstract
In this paper, we present GradPIM, a processing-in-memory architecture which accelerates parameter updates of deep neural networks training. As one of processing-in-memory techniques that could be realized in the near future, we propose an incremental, simple architectural design that does not invade the existing memory protocol. Extending DDR4 SDRAM to utilize bank-group parallelism makes our operation designs in processing-in-memory (PIM) module efficient in terms of hardware cost and performance. Our experimental results show that the proposed architecture can improve the performance of DNN training and greatly reduce memory bandwidth requirement while posing only a minimal amount of overhead to the protocol and DRAM area.
arXiv commentAccepted to HPCA 2021
References in corpus (5)
- SGDR: Stochastic Gradient Descent with Warm Restarts
- Horovod: fast and easy distributed deep learning in TensorFlow
- Highly Scalable Deep Learning Training System with Mixed-Precision: Training ImageNet in Four Minutes
- Distributed Deep Learning Using Synchronous Stochastic Gradient Descent
- Mixed Precision Training of Convolutional Neural Networks using Integer Operations