{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,10,4]],"date-time":"2025-10-04T07:52:18Z","timestamp":1759564338806,"version":"3.37.3"},"reference-count":50,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"9","license":[{"start":{"date-parts":[[2024,9,1]],"date-time":"2024-09-01T00:00:00Z","timestamp":1725148800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2024,9,1]],"date-time":"2024-09-01T00:00:00Z","timestamp":1725148800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2024,9,1]],"date-time":"2024-09-01T00:00:00Z","timestamp":1725148800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"Hong Kong RGC","award":["HKU 17204423","C7004-22G"],"award-info":[{"award-number":["HKU 17204423","C7004-22G"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Parallel Distrib. Syst."],"published-print":{"date-parts":[[2024,9]]},"DOI":"10.1109\/tpds.2024.3429625","type":"journal-article","created":{"date-parts":[[2024,7,18]],"date-time":"2024-07-18T17:41:46Z","timestamp":1721324506000},"page":"1644-1656","source":"Crossref","is-referenced-by-count":1,"title":["Swift: Expedited Failure Recovery for Large-Scale DNN Training"],"prefix":"10.1109","volume":"35","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-3859-1566","authenticated-orcid":false,"given":"Yuchen","family":"Zhong","sequence":"first","affiliation":[{"name":"University of Hong Kong, Hong Kong, SAR, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3395-3994","authenticated-orcid":false,"given":"Guangming","family":"Sheng","sequence":"additional","affiliation":[{"name":"University of Hong Kong, Hong Kong, SAR, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Juncheng","family":"Liu","sequence":"additional","affiliation":[{"name":"OneFlow Inc., Beijing, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jinhui","family":"Yuan","sequence":"additional","affiliation":[{"name":"OneFlow Inc., Beijing, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3144-4398","authenticated-orcid":false,"given":"Chuan","family":"Wu","sequence":"additional","affiliation":[{"name":"University of Hong Kong, Hong Kong, SAR, China"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1810.04805"},{"article-title":"Megatron-LM: Training multi-billion parameter language models using model parallelism","year":"2019","author":"Shoeybi","key":"ref2"},{"article-title":"Language models are few-shot learners","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Brown","key":"ref3"},{"key":"ref4","first-page":"947","article-title":"Analysis of large-scale multi-tenant GPU clusters for DNN training workloads","volume-title":"Proc. USENIX Annu. Tech. Conf.","author":"Jeon"},{"key":"ref5","doi-asserted-by":"crossref","DOI":"10.1145\/3458817.3476209","article-title":"Efficient large-scale language model training on GPU clusters using megatron-LM","volume-title":"Proc. Int. Conf. High Perform. Comput. Netw. Storage Anal.","author":"Narayanan"},{"key":"ref6","first-page":"637","article-title":"Understanding and improving failure tolerant training for deep learning recommendation with partial recovery","volume-title":"Proc. 4th Conf. Mach. Learn. Syst.","author":"Maeng"},{"key":"ref7","first-page":"929","article-title":"Check-n-Run: A checkpointing system for training deep learning recommendation models","volume-title":"Proc. 19th USENIX Symp. Netw. Syst. Des. Implementation","author":"Eisenman"},{"key":"ref8","first-page":"265","article-title":"TensorFlow: A system for large-scale machine learning","volume-title":"Proc. 12th USENIX Symp. Operating Syst. Des. Implementation","author":"Abadi"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.5555\/3454287.3455008"},{"key":"ref10","first-page":"203","article-title":"CheckFreq: Frequent, fine-grained DNN checkpointing","volume-title":"Proc. 19th USENIX Conf. File Storage Technol.","author":"Mohan"},{"year":"2022","key":"ref11","article-title":"Elastic horovod"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1145\/3437801.3441593"},{"key":"ref13","first-page":"15","article-title":"Resilient distributed datasets: A fault-tolerant abstraction for in-memory cluster computing","volume-title":"Proc. 9th USENIX Symp. Netw. Syst. Des. Implementation","author":"Zaharia"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1145\/3341301.3359653"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/ICDE.2005.72"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.14778\/2735496.2735506"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1145\/3341301.3359646"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.5244\/C.30.87"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2010.11929"},{"article-title":"Revisiting distributed synchronous SGD","year":"2016","author":"Chen","key":"ref20"},{"article-title":"Accurate, large minibatch SGD: Training ImageNet in 1 hour","year":"2017","author":"Goyal","key":"ref21"},{"article-title":"Large batch optimization for deep learning: Training BERT in 76 minutes","volume-title":"Proc. Int. Conf. Learn. Representations","author":"You","key":"ref22"},{"key":"ref23","first-page":"1","article-title":"Beyond data and model parallelism for deep neural networks","volume-title":"Proc. 2nd Conf. Mach. Learn. Syst.","author":"Jia"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.48550\/arxiv.1811.06965"},{"key":"ref25","first-page":"7937","article-title":"Memory-efficient pipeline-parallel DNN training","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Narayanan"},{"article-title":"OneFlow: Redesign the distributed deep learning framework from scratch","year":"2021","author":"Yuan","key":"ref26"},{"key":"ref27","first-page":"559","article-title":"Alpa: Automating inter- and intra-operator parallelism for distributed deep learning","volume-title":"Proc. 16th USENIX Symp. Operating Syst. Des. Implementation","author":"Zheng"},{"key":"ref28","first-page":"550","article-title":"A distributed graph-theoretic framework for automatic parallelization in multi-core systems","volume-title":"Proc. Mach. Learn. Syst.","author":"Ma"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/TVLSI.2019.2897650"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1109\/TC.2021.3071507"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1145\/3458817.3476205"},{"article-title":"Adam: A method for stochastic optimization","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Kingma","key":"ref32"},{"article-title":"MXNet: A flexible and efficient machine learning library for heterogeneous distributed systems","year":"2015","author":"Chen","key":"ref34"},{"key":"ref35","first-page":"5220","article-title":"Fault tolerance in iterative-convergent machine learning","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Qiao"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.5555\/3291168.3291210"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1145\/357369.357371"},{"article-title":"Decoupled weight decay regularization","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Loshchilov","key":"ref38"},{"article-title":"On the convergence of adam and beyond","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Reddi","key":"ref39"},{"volume-title":"Invertibility and Singularity for Bounded Linear Operators","year":"2016","author":"Harte","key":"ref40"},{"key":"ref41","first-page":"2214","article-title":"The reversible residual network: Backpropagation without storing activations","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Gomez"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1109\/MSST.2010.5496972"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO.2016.7783721"},{"article-title":"Optimized inter-GPU collective operations with NCCL 2","year":"2017","author":"Sylvain","key":"ref44"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-015-0816-y"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46493-0_38"},{"key":"ref47","first-page":"937","article-title":"KungFu: Making training in distributed machine learning adaptive","volume-title":"Proc. 14th USENIX Symp. Operating Syst. Des. Implementation","author":"Mai"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1145\/3190508.3190517"},{"key":"ref49","first-page":"1","article-title":"Pollux: Co-adaptive cluster scheduling for goodput-optimized deep learning","volume-title":"Proc. 15th USENIX Symp. Operating Syst. Des. Implementation","author":"Qiao"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1145\/3267809.3267810"},{"article-title":"Mixed precision training","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Micikevicius","key":"ref51"}],"container-title":["IEEE Transactions on Parallel and Distributed Systems"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/71\/10601540\/10601499.pdf?arnumber=10601499","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,8,3]],"date-time":"2024-08-03T04:40:28Z","timestamp":1722660028000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/10601499\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,9]]},"references-count":50,"journal-issue":{"issue":"9"},"URL":"https:\/\/doi.org\/10.1109\/tpds.2024.3429625","relation":{},"ISSN":["1045-9219","1558-2183","2161-9883"],"issn-type":[{"type":"print","value":"1045-9219"},{"type":"electronic","value":"1558-2183"},{"type":"electronic","value":"2161-9883"}],"subject":[],"published":{"date-parts":[[2024,9]]}}}