{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,30]],"date-time":"2026-10-30T02:29:15Z","timestamp":1793327355559,"version":"4.5.0"},"reference-count":32,"publisher":"IEEE","license":[{"start":{"date-parts":[[2020,11,1]],"date-time":"2020-11-01T00:00:00Z","timestamp":1604188800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2020,11,1]],"date-time":"2020-11-01T00:00:00Z","timestamp":1604188800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2020,11,1]],"date-time":"2020-11-01T00:00:00Z","timestamp":1604188800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2020,11]]},"DOI":"10.1109\/sc41405.2020.00024","type":"proceedings-article","created":{"date-parts":[[2021,2,22]],"date-time":"2021-02-22T20:26:48Z","timestamp":1614025608000},"page":"1-16","source":"Crossref","is-referenced-by-count":790,"title":["ZeRO: Memory optimizations Toward Training Trillion Parameter Models"],"prefix":"10.1109","author":[{"given":"Samyam","family":"Rajbhandari","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jeff","family":"Rasley","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Olatunji","family":"Ruwase","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yuxiong","family":"He","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref32","author":"kaplan","year":"2020","journal-title":"Scaling laws for neural language models"},{"key":"ref31","article-title":"Roberta: A robustly optimized BERT pretraining approach","volume":"abs 1907 11692","author":"liu","year":"2019","journal-title":"CoRR"},{"key":"ref30","author":"narasimhan","year":"2019","journal-title":"NVIDIA Clocks World’s Fastest BERT Training Time"},{"key":"ref10","year":"2020","journal-title":"Turing-NLG A 17-Billion-Parameter Language Model by Microsoft"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1145\/3302424.3303953"},{"key":"ref12","article-title":"Gpipe: Efficient training of giant neural networks using pipeline parallelism","volume":"abs 1811 6965","author":"huang","year":"2018","journal-title":"ArXiv"},{"key":"ref13","article-title":"Pipedream: Fast and efficient pipeline parallel DNN training","volume":"abs 1806 3377","author":"harlap","year":"2018","journal-title":"CoRR"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1145\/3341301.3359646"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA.2018.00070"},{"key":"ref16","article-title":"Checkmate: Breaking the memory wall with optimal tensor rematerialization","volume":"abs 1910 2653","author":"jain","year":"2019","journal-title":"ArXiv"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1145\/3178487.3178491"},{"key":"ref18","article-title":"Training large neural networks with constant memory using a new execution algorithm","volume":"abs 2002 5645","author":"pudipeddi","year":"2020","journal-title":"ArXiv"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO.2016.7783721"},{"key":"ref28","year":"2019","journal-title":"Automatic mixed precision (amp)"},{"key":"ref4","author":"shoeybi","year":"2019","journal-title":"Megatron-lm Training multi-billion parameter language models using model parallelism"},{"key":"ref27","year":"2017","journal-title":"Nvidia tesla v100 gpu architecture"},{"key":"ref3","author":"radford","year":"2019","journal-title":"Language Models are Unsupervised Multitask Learners"},{"key":"ref6","article-title":"Adam: A method for stochastic optimization","author":"diederik","year":"2015","journal-title":"3rd International Conference on Learning Representations ICLR 2015"},{"key":"ref29","article-title":"Horovod: fast and easy distributed deep learning in tensorflow","author":"sergeev","year":"2018","journal-title":"arXiv preprint arxiv 1802 05807"},{"key":"ref5","author":"raffel","year":"2019","journal-title":"Exploring the limits of transfer learning with a unified text-to-text transformer"},{"key":"ref8","article-title":"Training deep nets with sublinear memory cost","volume":"abs 1604 6174","author":"chen","year":"2016","journal-title":"CoRR"},{"key":"ref7","article-title":"Mesh-tensorflow: Deep learning for supercomputers","volume":"abs 1811 2084","author":"shazeer","year":"2018","journal-title":"CoRR"},{"key":"ref2","article-title":"BERT: pre-training of deep bidirectional transformers for language understanding","volume":"abs 1810 4805","author":"devlin","year":"2018","journal-title":"CoRR"},{"key":"ref9","article-title":"An empirical model of large-batch training","volume":"abs 1812 6162","author":"mccandlish","year":"2018","journal-title":"CoRR"},{"key":"ref1","article-title":"Attention is all you need","volume":"abs 1706 3762","author":"vaswani","year":"2017","journal-title":"CoRR"},{"key":"ref20","article-title":"TFLMS: large model support in tensorflow by graph rewriting","volume":"abs 1807 2037","author":"le","year":"2018","journal-title":"CoRR"},{"key":"ref22","article-title":"Memoryefficient adaptive optimization for large-scale learning","volume":"abs 1901 11150","author":"anil","year":"2019","journal-title":"ArXiv"},{"key":"ref21","article-title":"Adafactor: Adaptive learning rates with sublinear memory cost","volume":"abs 1804 4235","author":"shazeer","year":"2018","journal-title":"CoRR"},{"key":"ref24","article-title":"Scaling SGD batch size to 32k for imagenet training","volume":"abs 1708 3888","author":"you","year":"2017","journal-title":"CoRR"},{"key":"ref23","first-page":"2121","article-title":"Adaptive subgradient methods for online learning and stochastic optimization","volume":"12","author":"duchi","year":"2011","journal-title":"J Mach Learn Res"},{"key":"ref26","author":"micikevicius","year":"2017","journal-title":"Mixed Precision Training"},{"key":"ref25","article-title":"Reducing BERT pre-training time from 3 days to 76 minutes","volume":"abs 1904 962","author":"you","year":"2019","journal-title":"CoRR"}],"event":{"name":"SC20: International Conference for High Performance Computing, Networking, Storage and Analysis","location":"Atlanta, GA, USA","start":{"date-parts":[[2020,11,9]]},"end":{"date-parts":[[2020,11,19]]}},"container-title":["SC20: International Conference for High Performance Computing, Networking, Storage and Analysis"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/9355221\/9355202\/09355301.pdf?arnumber=9355301","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2022,6,28]],"date-time":"2022-06-28T17:56:26Z","timestamp":1656438986000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/9355301\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2020,11]]},"references-count":32,"URL":"https:\/\/doi.org\/10.1109\/sc41405.2020.00024","relation":{},"subject":[],"published":{"date-parts":[[2020,11]]}}}