| Zero: Memory optimizations toward training trillion parameter models S Rajbhandari, J Rasley, O Ruwase, Y He SC20: International Conference for High Performance Computing, Networking …, 2020 | 490 | 2020 |
| Deepspeed: System optimizations enable training deep learning models with over 100 billion parameters J Rasley, S Rajbhandari, O Ruwase, Y He Proceedings of the 26th ACM SIGKDD International Conference on Knowledge …, 2020 | 382 | 2020 |
| Using deepspeed and megatron to train megatron-turing nlg 530b, a large-scale generative language model S Smith, M Patwary, B Norick, P LeGresley, S Rajbhandari, J Casper, ... arXiv preprint arXiv:2201.11990, 2022 | 315 | 2022 |
| Graph query processing using plurality of engines S Elnikety, Y He, S Sakr US Patent 9,053,210, 2015 | 283 | 2015 |
| {ZeRO-Offload}: Democratizing {Billion-Scale} model training J Ren, S Rajbhandari, RY Aminabadi, O Ruwase, S Yang, M Zhang, D Li, ... 2021 USENIX Annual Technical Conference (USENIX ATC 21), 551-564, 2021 | 155 | 2021 |
| Provably-efficient job scheduling for energy and fairness in geographically distributed data centers S Ren, Y He, F Xu 2012 IEEE 32nd International Conference on Distributed Computing Systems, 22-31, 2012 | 152 | 2012 |
| Learning intrinsic sparse structures within long short-term memory W Wen, Y He, S Rajbhandari, M Zhang, W Wang, F Liu, B Hu, Y Chen, ... arXiv preprint arXiv:1709.05027, 2017 | 147 | 2017 |
| The Cilkview scalability analyzer Y He, CE Leiserson, WM Leiserson Proceedings of the twenty-second annual ACM symposium on Parallelism in …, 2010 | 141 | 2010 |
| Adaptive work-stealing with parallelism feedback K Agrawal, CE Leiserson, Y He, WJ Hsu ACM Transactions on Computer Systems (TOCS) 26 (3), 1-32, 2008 | 138 | 2008 |
| Zero-infinity: Breaking the gpu memory wall for extreme scale deep learning S Rajbhandari, O Ruwase, J Rasley, S Smith, Y He Proceedings of the International Conference for High Performance Computing …, 2021 | 137 | 2021 |
| Few-to-many: Incremental parallelism for reducing tail latency in interactive services ME Haque, YH Eom, Y He, S Elnikety, R Bianchini, KS McKinley ACM SIGPLAN Notices 50 (4), 161-175, 2015 | 129 | 2015 |
| Predictive parallelization: Taming tail latencies in web search M Jeon, S Kim, S Hwang, Y He, S Elnikety, AL Cox, S Rixner Proceedings of the 37th international ACM SIGIR conference on Research …, 2014 | 118 | 2014 |
| Swayam: distributed autoscaling to meet slas of machine learning inference services with resource efficiency A Gujarati, S Elnikety, Y He, KS McKinley, BB Brandenburg Proceedings of the 18th ACM/IFIP/USENIX middleware conference, 109-120, 2017 | 111 | 2017 |
| {DeepCPU}: Serving {RNN-based} Deep Learning Models 10x Faster M Zhang, S Rajbhandari, W Wang, Y He 2018 USENIX Annual Technical Conference (USENIX ATC 18), 951-965, 2018 | 108 | 2018 |
| Performance modeling and scalability optimization of distributed deep learning systems F Yan, O Ruwase, Y He, T Chilimbi Proceedings of the 21th ACM SIGKDD International Conference on Knowledge …, 2015 | 97 | 2015 |
| Adaptive scheduling with parallelism feedback K Agrawal, Y He, WJ Hsu, CE Leiserson Proceedings of the eleventh ACM SIGPLAN symposium on Principles and practice …, 2006 | 93 | 2006 |
| Zeta: Scheduling interactive services with partial execution Y He, S Elnikety, J Larus, C Yan Proceedings of the Third ACM Symposium on Cloud Computing, 1-14, 2012 | 83 | 2012 |
| Deepspeed-moe: Advancing mixture-of-experts inference and training to power next-generation ai scale S Rajbhandari, C Li, Z Yao, M Zhang, RY Aminabadi, AA Awan, J Rasley, ... International Conference on Machine Learning, 18332-18346, 2022 | 79 | 2022 |
| Mercury: A memory-constrained spatio-temporal real-time search on microblogs A Magdy, MF Mokbel, S Elnikety, S Nath, Y He 2014 IEEE 30th International Conference on Data Engineering, 172-183, 2014 | 78 | 2014 |
| Exploiting heterogeneity for tail latency and energy efficiency ME Haque, Y He, S Elnikety, TD Nguyen, R Bianchini, KS McKinley Proceedings of the 50th Annual IEEE/ACM International Symposium on …, 2017 | 75 | 2017 |
