{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,4]],"date-time":"2026-06-04T12:26:46Z","timestamp":1780576006547,"version":"3.54.1"},"reference-count":52,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"5","license":[{"start":{"date-parts":[[2023,5,1]],"date-time":"2023-05-01T00:00:00Z","timestamp":1682899200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2023,5,1]],"date-time":"2023-05-01T00:00:00Z","timestamp":1682899200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2023,5,1]],"date-time":"2023-05-01T00:00:00Z","timestamp":1682899200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"National Key R&D Program of China","award":["2021YFB0301200"],"award-info":[{"award-number":["2021YFB0301200"]}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["62025208"],"award-info":[{"award-number":["62025208"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Parallel Distrib. Syst."],"published-print":{"date-parts":[[2023,5]]},"DOI":"10.1109\/tpds.2023.3247001","type":"journal-article","created":{"date-parts":[[2023,3,17]],"date-time":"2023-03-17T17:25:29Z","timestamp":1679073929000},"page":"1466-1478","source":"Crossref","is-referenced-by-count":59,"title":["Merak: An Efficient Distributed DNN Training Framework With Automated 3D Parallelism for Giant Foundation Models"],"prefix":"10.1109","volume":"34","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-3458-4732","authenticated-orcid":false,"given":"Zhiquan","family":"Lai","sequence":"first","affiliation":[{"name":"National Laboratory for Parallel and Distributed Processing, College of Computer, National University of Defense Technology, Changsha, Hunan, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7419-1511","authenticated-orcid":false,"given":"Shengwei","family":"Li","sequence":"additional","affiliation":[{"name":"National Laboratory for Parallel and Distributed Processing, College of Computer, National University of Defense Technology, Changsha, Hunan, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xudong","family":"Tang","sequence":"additional","affiliation":[{"name":"National Laboratory for Parallel and Distributed Processing, College of Computer, National University of Defense Technology, Changsha, Hunan, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0669-6892","authenticated-orcid":false,"given":"Keshi","family":"Ge","sequence":"additional","affiliation":[{"name":"National Laboratory for Parallel and Distributed Processing, College of Computer, National University of Defense Technology, Changsha, Hunan, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0354-9467","authenticated-orcid":false,"given":"Weijie","family":"Liu","sequence":"additional","affiliation":[{"name":"National Laboratory for Parallel and Distributed Processing, College of Computer, National University of Defense Technology, Changsha, Hunan, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yabo","family":"Duan","sequence":"additional","affiliation":[{"name":"National Laboratory for Parallel and Distributed Processing, College of Computer, National University of Defense Technology, Changsha, Hunan, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8285-2738","authenticated-orcid":false,"given":"Linbo","family":"Qiao","sequence":"additional","affiliation":[{"name":"National Laboratory for Parallel and Distributed Processing, College of Computer, National University of Defense Technology, Changsha, Hunan, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9743-2034","authenticated-orcid":false,"given":"Dongsheng","family":"Li","sequence":"additional","affiliation":[{"name":"National Laboratory for Parallel and Distributed Processing, College of Computer, National University of Defense Technology, Changsha, Hunan, China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"On the opportunities and risks of foundation models","author":"Bommasani","year":"2021"},{"key":"ref2","article-title":"Attention is all you need","volume-title":"Adv. Neural Inf. Process. Syst.","volume":"30","author":"Vaswani","year":"2017"},{"key":"ref3","first-page":"1877","article-title":"Language models are few-shot learners","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Brown"},{"key":"ref4","article-title":"Using deepspeed and megatron to train megatron-turing NLG 530B, a large-scale generative language model","author":"Smith","year":"2022"},{"key":"ref5","article-title":"Pangu-: Large-scale autoregressive pretrained chinese language models with auto-parallel computation","author":"Zeng","year":"2021"},{"key":"ref6","article-title":"DeepSpeed: Extreme-scale model training for everyone","year":"2020"},{"key":"ref7","first-page":"1","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":"ref8","article-title":"Colossal-AI: A unified deep learning system for large-scale parallel training","author":"Bian","year":"2021"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1145\/3492321.3519584"},{"key":"ref10","article-title":"Amazon sagemaker model parallelism: A general and flexible framework for large model training","author":"Karakus","year":"2021"},{"key":"ref11","article-title":"Oneflow: Redesign the distributed deep learning framework from scratch","author":"Yuan","year":"2022"},{"key":"ref12","article-title":"End-to-end adaptive distributed training on paddlepaddle","author":"Ao","year":"2021"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.5555\/3454287.3455008"},{"key":"ref14","first-page":"38","article-title":"Transformers: State-of-the-art natural language processing","author":"Wolf","year":"2020"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1145\/3394486.3406703"},{"key":"ref16","article-title":"Megatron-LM: Training multi-billion parameter language models using model parallelism","author":"Shoeybi","year":"2019"},{"issue":"8","key":"ref17","article-title":"Language models are unsupervised multitask learners","volume":"1","author":"Radford","year":"2019","journal-title":"OpenAI Blog"},{"key":"ref18","article-title":"Exploring the limits of transfer learning with a unified text-to-text transformer","author":"Raffel","year":"2020"},{"key":"ref19","article-title":"BERT: Pre-training of deep bidirectional transformers for language understanding","author":"Devlin","year":"2018"},{"key":"ref20","first-page":"673","article-title":"Whale: Efficient giant model training over heterogeneous GPUs","volume-title":"Proc. USENIX Annu. Tech. Conf.","author":"Jia"},{"key":"ref21","article-title":"Horovod: Fast and easy distributed deep learning in tensorflow","author":"Sergeev","year":"2018"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.14778\/3415478.3415530"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.14778\/3503585.3503590"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1016\/j.jpdc.2008.09.002"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1109\/SC41405.2020.00024"},{"key":"ref26","first-page":"551","article-title":"ZeRO-Offload: Democratizing billion-scale model training","volume-title":"Proc. USENIX Annu. Tech. Conf.","author":"Ren"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1145\/3458817.3476205"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/tpds.2022.3219819"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1145\/3341301.3359646"},{"key":"ref30","first-page":"7937","article-title":"Memory-efficient pipeline-parallel DNN training","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Narayanan"},{"key":"ref31","first-page":"307","article-title":"HetPipe: Enabling large DNN training on (whimpy) heterogeneous GPU clusters through integration of pipelined model parallelism and data parallelism","volume-title":"Proc. USENIX Annu. Techn. Conf.","author":"Park"},{"key":"ref32","first-page":"381","article-title":"Fine-tuning giant neural networks on commodity hardware with automatic pipeline model parallelism","volume-title":"Proc. USENIX Annu. Tech. Conf.","author":"Eliad"},{"key":"ref33","first-page":"479","article-title":"Pipelined backpropagation at scale: Training large models without batches","volume-title":"Proc. Mach. Learn. Syst. Conf.","author":"Kosson"},{"key":"ref34","first-page":"269","article-title":"PipeMare: Asynchronous pipeline parallel DNN training","volume-title":"Proc. Mach. Learn. Syst. Conf.","author":"Yang"},{"key":"ref35","article-title":"GPipe: Efficient training of giant neural networks using pipeline parallelism","volume-title":"Adv. Neural Inf. Process. Syst.","volume":"32","author":"Huang","year":"2019"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/CBD.2019.00020"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1145\/3437801.3441593"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1145\/3472456.3472497"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1145\/3458817.3476145"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1109\/ipdps54959.2023.00031"},{"key":"ref41","article-title":"2.5-dimensional distributed model training","author":"Wang","year":"2021"},{"key":"ref42","article-title":"Maximizing parallelism in distributed training for huge neural networks","author":"Bian","year":"2021"},{"key":"ref43","article-title":"Training deep nets with sublinear memory cost","author":"Chen","year":"2016"},{"key":"ref44","article-title":"Dynamic tensor rematerialization","author":"Kirisame","year":"2020"},{"key":"ref45","first-page":"497","article-title":"Checkmate: Breaking the memory wall with optimal tensor rematerialization","volume-title":"Proc. Conf. Mach. Learn. Syst.","author":"Jain"},{"key":"ref46","article-title":"A survey on auto-parallelism of neural networks training","author":"Liang","year":"2022"},{"key":"ref47","first-page":"638","article-title":"Torch.fx: Practical program capture and transformation for deep learning in python","volume-title":"Proc. Conf. Mach. Learn. Syst.","author":"Reed"},{"key":"ref48","article-title":"An image is worth 16x16 words: Transformers for image recognition at scale","author":"Dosovitskiy","year":"2020"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00986"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1109\/CLUSTER51413.2022.00042"},{"key":"ref51","article-title":"NVIDIA collective communications library (NCCL)","year":"2019"},{"key":"ref52","article-title":"Openwebtext corpus","author":"Gokaslan","year":"2019"}],"container-title":["IEEE Transactions on Parallel and Distributed Systems"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/71\/10075651\/10049507.pdf?arnumber=10049507","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,2,13]],"date-time":"2024-02-13T15:45:07Z","timestamp":1707839107000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/10049507\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2023,5]]},"references-count":52,"journal-issue":{"issue":"5"},"URL":"https:\/\/doi.org\/10.1109\/tpds.2023.3247001","relation":{},"ISSN":["1045-9219","1558-2183","2161-9883"],"issn-type":[{"value":"1045-9219","type":"print"},{"value":"1558-2183","type":"electronic"},{"value":"2161-9883","type":"electronic"}],"subject":[],"published":{"date-parts":[[2023,5]]}}}