A Neural Network-Based Pipeline Parallel Strategy Solver for Heterogeneous Environments

A Neural Network-Based Pipeline Parallel Strategy Solver for Heterogeneous Environments | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article A Neural Network-Based Pipeline Parallel Strategy Solver for Heterogeneous Environments Jie Ou, Yueming Chen, Wenhong Tian, Hongjie Zhang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4669385/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The escalating interest in language models has accentuated the significance of distributed training approaches, specifically pipeline parallelism. Most investigations primarily employ dynamic programming, these methods face inherent limitations in computational efficiency and do not offer a straightforward end-to-end resolution pathway. Furthermore, their applicability is significantly curtailed in heterogeneous environments. Addressing this challenge, we introduce a novel Neural Network-based Pipeline Parallel strategy solver (NNPiper) tailored for heterogeneous environments. NNPiper can perceive computational and communication costs, the number of stages to be divided, and the number of micro-batches, and it can directly provide the strategy for allocating specific devices to each pipeline stage, without requiring online training based on actual cluster data. Our unique neural network model and training method, the Virtual Contrastive Training Algorithm (VCTA), enables efficient training of NNPiper without collecting large amounts of real data. Compared with the state-of-the-art method, NNPiper can improve the theoretical training speed on average by 17.975%, in the completely heterogeneous environment for the Generative Pre-trained Transformer (GPT) models. pipeline parallel strategy heterogeneous environments neural network-based strategy virtual contrastive training algorithm training speed Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4669385","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":324954470,"identity":"8333ea1e-9a37-46dd-8154-5e0016a95353","order_by":0,"name":"Jie Ou","email":"","orcid":"","institution":"University of Electronic Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Jie","middleName":"","lastName":"Ou","suffix":""},{"id":324954471,"identity":"93a4518a-fb1c-46da-9174-b519f336c5da","order_by":1,"name":"Yueming Chen","email":"","orcid":"","institution":"University of Electronic Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Yueming","middleName":"","lastName":"Chen","suffix":""},{"id":324954472,"identity":"dc02e605-c1e1-4dcd-ad39-00db41bf0a18","order_by":2,"name":"Wenhong Tian","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABBklEQVRIiWNgGAWjYDACZihtACI+wER58OjgQdbCOAOmGq8WGAOkhZmHGC327MzPHjC22eSZs/cefm3bZmNvz36A8cHbNgZ5c5wOYzM3YGxLK7bsOZdmnduWltjDk8BsOLeNwXBnA06/mEkwbjucuOFGjplxbtvhBB4JBjZp3jaGBIMDuLSwfwNq+Q/RYtn23x6ohf03fi08IFsOgLQYP2ZsO8DYA7SFGa+WwzxlEoz/khM3nDljxthzLjmx50xis+SccxKGG3BoYe8/vk2C4Yxd4objPcYffpTZ2bO3Hz744U2ZjTwuW0CA+Q+EZpOA0IwNQEICt3pkrR8IqxkFo2AUjIKRCADibFGB/e1bOwAAAABJRU5ErkJggg==","orcid":"","institution":"University of Electronic Science and Technology of China","correspondingAuthor":true,"prefix":"","firstName":"Wenhong","middleName":"","lastName":"Tian","suffix":""},{"id":324954473,"identity":"e62a6d91-863b-4572-a7bb-dec620338a63","order_by":3,"name":"Hongjie Zhang","email":"","orcid":"","institution":"Sichuan Normal University","correspondingAuthor":false,"prefix":"","firstName":"Hongjie","middleName":"","lastName":"Zhang","suffix":""}],"badges":[],"createdAt":"2024-07-01 16:25:21","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4669385/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4669385/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":73708707,"identity":"4e493e44-9a2d-4a74-a744-35d95197053b","added_by":"auto","created_at":"2025-01-13 19:46:37","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":7444114,"visible":true,"origin":"","legend":"","description":"","filename":"NNPiper240701.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4669385/v1_covered_0f6ef6f9-edab-4234-9f45-925beb272451.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Neural Network-Based Pipeline Parallel Strategy Solver for Heterogeneous Environments","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"pipeline parallel strategy, heterogeneous environments, neural network-based strategy, virtual contrastive training algorithm, training speed","lastPublishedDoi":"10.21203/rs.3.rs-4669385/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4669385/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe escalating interest in language models has accentuated the significance of distributed training approaches, specifically pipeline parallelism. Most investigations primarily employ dynamic programming, these methods face inherent limitations in computational efficiency and do not offer a straightforward end-to-end resolution pathway. Furthermore, their applicability is significantly curtailed in heterogeneous environments. Addressing this challenge, we introduce a novel Neural Network-based Pipeline Parallel strategy solver (NNPiper) tailored for heterogeneous environments. NNPiper can perceive computational and communication costs, the number of stages to be divided, and the number of micro-batches, and it can directly provide the strategy for allocating specific devices to each pipeline stage, without requiring online training based on actual cluster data. Our unique neural network model and training method, the Virtual Contrastive Training Algorithm (VCTA), enables efficient training of NNPiper without collecting large amounts of real data. Compared with the state-of-the-art method, NNPiper can improve the theoretical training speed on average by 17.975%, in the completely heterogeneous environment for the Generative Pre-trained Transformer (GPT) models.\u003c/p\u003e","manuscriptTitle":"A Neural Network-Based Pipeline Parallel Strategy Solver for Heterogeneous Environments","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-07-24 04:52:48","doi":"10.21203/rs.3.rs-4669385/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"dd4ada42-dc92-4822-a2ae-74c50f511d6f","owner":[],"postedDate":"July 24th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-01-13T19:38:20+00:00","versionOfRecord":[],"versionCreatedAt":"2024-07-24 04:52:48","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4669385","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4669385","identity":"rs-4669385","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}