HyPar: Towards Hybrid Parallelism for Deep Learning Accelerator Array
Open accessThe DOI matches its Crossref record; 15 of the 15 references it cites were found in a registry or an index.
Abstract
With the rise of artificial intelligence in recent years, Deep Neural Networks (DNNs) have been widely used in many domains. To achieve high performance and energy efficiency, hardware acceleration (especially inference) of DNNs is intensively studied both in academia and industry. However, we still face two challenges: large DNN models and datasets, which incur frequent off-chip memory accesses; and the training of DNNs, which is not well-explored in recent accelerator designs. To truly provide high throughput and energy efficient acceleration for the training of deep and large models, we inevitably need to use multiple accelerators to explore the coarse-grain parallelism, compared to the fine-grain parallelism inside a layer considered in most of the existing architectures. It poses the key research question to seek the best organization of computation and dataflow among accelerators. In this paper, we propose a solution HYPAR to determine layer-wise parallelism for deep neural network training with an array of DNN accelerators. HYPAR partitions the feature map tensors (input and output), the kernel tensors, the gradient tensors, and the error tensors for the DNN accelerators. A partition constitutes the choice of parallelism for weighted layers. The optimization target is to search a partition that minimizes the total communication during training a complete DNN. To solve this problem, we propose a communication model to explain the source and amount of communications. Then, we use a hierarchical layer-wise dynamic programming method to search for the partition for each layer. HYPAR is practical: the time complexity for the partition search in HYPAR is linear. We apply this method in an HMC-based DNN training architecture to minimize data movement. We evaluate HYPAR with ten DNN models from classic Lenet to large-size model VGGs, and the number of weighted layers of these models range from four to nineteen. Our evaluation finds that: the default Model Parallelism is indeed the worst; the default Data Parallelism is not the best; but hybrid parallelism can be better than either the default Data Parallelism or Model Parallelism in DNN training with an array of accelerators. Our evaluation shows that HYPAR achieves a performance gain of 3.39× and an energy efficiency gain of 1.51× compared to Data Parallelism on average, and HYPAR performs up to 2.40× better than “one weird trick”.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
ImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333
In-Datacenter Performance Analysis of a Tensor Processing Unit
Jouppi, N. P., Young, C., Patil, N., Patterson, D., Agrawal, G., Bajwa, R., Bates, S., Bhatia, S., Boden, N., Borchers, A., et al. (2017) · Proceedings of the 44th Annual International Symposium on Computer Architecture · doi:10.1145/3079856.3080246 · Cited by 4,537
Eyeriss: An Energy-Efficient Reconfigurable Accelerator for Deep Convolutional Neural Networks
Chen, Y.-H., Krishna, T., Emer, J. S., & Sze, V. (2017) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2016.2616357 · Cited by 3,247
Self-citationLearning Hierarchical Features for Scene Labeling
Farabet, C., Couprie, C., Najman, L., & LeCun, Y. (2013) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2012.231 · Cited by 2,620
Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour
Goyal, P., Dollár, P., Girshick, R., Noordhuis, P., Wesolowski, L., Kyrola, A., Tulloch, A., Jia, Y., & He, K. (2017) · arXiv · doi:10.48550/arxiv.1706.02677 · Cited by 2,294
Optimizing FPGA-based Accelerator Design for Deep Convolutional Neural Networks
Zhang, C., Li, P., Sun, G., Guan, Y., Xiao, B., & Cong, J. (2015) · Proceedings of the 2015 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays · doi:10.1145/2684746.2689060 · Cited by 2,158
Eyeriss: a spatial architecture for energy-efficient dataflow for convolutional neural networks
Chen, Y.-H., Emer, J., & Sze, V. (2016) · ACM SIGARCH Computer Architecture News · doi:10.1145/3007787.3001177 · Cited by 1,182
Self-citationEIE: Efficient Inference Engine on Compressed Deep Neural Network
Han, S., Liu, X., Mao, H., Pu, J., Pedram, A., Horowitz, M. A., & Dally, W. J. (2016) · 2016 ACM/IEEE 43rd Annual International Symposium on Computer Architecture (ISCA) · doi:10.1109/isca.2016.30 · Cited by 1,077
Alwani, M., Chen, H., Ferdman, M., & Milder, P. (2016) · 2016 49th Annual IEEE/ACM International Symposium on Microarchitecture (MICRO) · doi:10.1109/micro.2016.7783725 · Cited by 662
ESE: Efficient Speech Recognition Engine with Sparse LSTM on FPGA
Han, S., Kang, J., Mao, H., Hu, Y., Li, X., Li, Y., Xie, D., Luo, H., Yao, S., Wang, Y., et al. (2017) · Proceedings of the 2017 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays · doi:10.1145/3020078.3021745 · Cited by 591
MoDNN: Local distributed mobile computing system for Deep Neural Network
Mao, J., Chen, X., Nixon, K. W., Krieger, C., & Chen, Y. (2017) · Design, Automation & Test in Europe Conference & Exhibition (DATE), 2017 · doi:10.23919/date.2017.7927211 · Cited by 331
Self-citationMeDNN: A distributed mobile system with enhanced partition and deployment for large-scale DNNs
Mao, J., Yang, Z., Wen, W., Wu, C., Song, L., Nixon, K. W., Chen, X., Li, H., & Chen, Y. (2017) · 2017 IEEE/ACM International Conference on Computer-Aided Design (ICCAD) · doi:10.1109/iccad.2017.8203852 · Cited by 97
Self-citation
Cited by
SIGMA: A Sparse and Irregular GEMM Accelerator with Flexible Interconnects for DNN Training
Qin, E., Samajdar, A., Kwon, H., Nadella, V., Srinivasan, S., Das, D., Kaul, B., & Krishna, T. (2020) · 2020 IEEE International Symposium on High Performance Computer Architecture (HPCA) · doi:10.1109/hpca47549.2020.00015 · Cited by 487
Efficient Acceleration of Deep Learning Inference on Resource-Constrained Edge Devices: A Review
Shuvo, M. M. H., Islam, S. K., Cheng, J., & Morshed, B. I. (2023) · Proceedings of the IEEE · doi:10.1109/jproc.2022.3226481 · Cited by 436
“Though accelerators such as ScaleDeep [419] and HyPar [420] are proposed for training on edge, these architectures do not consider the sparsity, variable precision of gradients, parameters, and activations.”
MethodA Survey of Accelerator Architectures for Deep Neural Networks
Chen, Y., Xie, Y., Song, L., Chen, F., & Tang, T. (2020) · Engineering · doi:10.1016/j.eng.2020.01.007 · Cited by 370
Self-citationYang, Q., Liu, Y., Cheng, Y., Kang, Y., Chen, T., & Yu, H. (2019) · Synthesis lectures on artificial intelligence and machine learning · doi:10.2200/s00960ed2v01y201910aim043 · Cited by 261
Adaptable butterfly accelerator for attention-based NNs via hardware and algorithm co-design
Fan, H., Chau, T., Venieris, S. I., Lee, R., Kouris, A., Luk, W., Lane, N. D., & Abdelfattah, M. S. (2022) · 2022 55th IEEE/ACM International Symposium on Microarchitecture (MICRO) · doi:10.1109/micro56248.2022.00050 · Cited by 66
Metadata
- Authors
- Song, L., Mao, J., Zhuo, Y., Qian, X., Li, H., & Chen, Y.
- Venue
- 2019 IEEE International Symposium on High Performance Computer Architecture (HPCA)
- Publisher
- IEEE
- Published
- 2019-02
- Pages
- 56-68
- ISSN
- 1530-0897 (ISSN-L)
- OpenAlex
- W2963358710
- Cited by
- 108
- References
- 15
- FWCI
- 14.10
- Type
- conference paper
- Language
- en
- Open access
- green
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.