Unified Efficient Fine-Tuning Techniques for Open-Source Large Language Models

Unified Efficient Fine-Tuning Techniques for Open-Source Large Language Models | 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 Unified Efficient Fine-Tuning Techniques for Open-Source Large Language Models Yulin Zhang, Yanhua Li, Junhan Liu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4660140/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 Natural language processing has been revolutionized through the development of powerful large language models capable of understanding and generating human-like text, yet fine-tuning these models for specific tasks remains a computationally demanding process. Our novel approach to efficient fine-tuning, utilizing techniques such as adapter layers, Low-Rank Adaptation (LoRA), and layer-wise freezing, offers a significant reduction in computational overhead while maintaining or enhancing model performance. Extensive experiments with the Llama model demonstrated notable improvements in accuracy, F1-score, precision, and recall across multiple tasks, alongside marked reductions in GPU usage, training time, and memory consumption. The fine-tuned model also exhibited enhanced adaptability and generalization capabilities, performing well on new and unseen tasks, thus proving the efficacy and practical benefits of our methods. The contributions of this research provide valuable insights into optimizing large language models, broadening their applicability, and making advanced natural language processing techniques more accessible and efficient. Artificial Intelligence and Machine Learning Fine-tuning Efficiency NLP Adapter Layers LoRA Adaptability Full Text Additional Declarations The authors declare no competing interests. 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-4660140","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":320622825,"identity":"35267205-9319-4462-b0ea-6ffe7de0af8c","order_by":0,"name":"Yulin Zhang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/0lEQVRIiWNgGAWjYBAC+xlAgrGNQYaNgf8Z84cKII8dKNCARwsjVAsPGwOPGbPEGQYGHmZitDCcYeBhAGnhbSNCC7N087OHXyoO8/BJ95g95p1nk7ifmfngwxkMdnK6OPSxyRwzN5YxOMzDJnP+uTHvtrTEHma2ZMMNDMnGZgewa+GRSDCTlgBpkchhk+bddhiohcdM8gHDgcRtOLRISKR/Q9IyhwgtBhI5ZpIf4FoaoFo24NdSJs1gkA7RInEszbjnMNAvMwxw+8V+Rvo2yR8G1nLyM4BaPtTYyLa3Nx982FNhJ4dLCwgw82CxHbdyEGD8gV9+FIyCUTAKRjoAAPjzUOqna8TsAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0009-0003-8925-126X","institution":"","correspondingAuthor":true,"prefix":"","firstName":"Yulin","middleName":"","lastName":"Zhang","suffix":""},{"id":320622826,"identity":"9dfcfcf3-544e-4a6c-a6e8-944249196710","order_by":1,"name":"Yanhua Li","email":"","orcid":"https://orcid.org/0009-0000-1513-9492","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Yanhua","middleName":"","lastName":"Li","suffix":""},{"id":320622827,"identity":"04f01788-a824-4a45-b5c7-e5cb92f11943","order_by":2,"name":"Junhan Liu","email":"","orcid":"https://orcid.org/0009-0000-4499-982X","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Junhan","middleName":"","lastName":"Liu","suffix":""}],"badges":[],"createdAt":"2024-06-29 16:35:29","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-4660140/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4660140/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":59454232,"identity":"6fa85963-be88-48f2-880c-2814758c57ee","added_by":"auto","created_at":"2024-07-02 03:10:24","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":177645,"visible":true,"origin":"","legend":"","description":"","filename":"RS.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4660140/v1_covered_0de9cb75-0a75-43cd-ba69-54a89e5f67bb.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eUnified Efficient Fine-Tuning Techniques for Open-Source Large Language Models\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"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":"Fine-tuning, Efficiency, NLP, Adapter Layers, LoRA, Adaptability","lastPublishedDoi":"10.21203/rs.3.rs-4660140/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4660140/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eNatural language processing has been revolutionized through the development of powerful large language models capable of understanding and generating human-like text, yet fine-tuning these models for specific tasks remains a computationally demanding process. Our novel approach to efficient fine-tuning, utilizing techniques such as adapter layers, Low-Rank Adaptation (LoRA), and layer-wise freezing, offers a significant reduction in computational overhead while maintaining or enhancing model performance. Extensive experiments with the Llama model demonstrated notable improvements in accuracy, F1-score, precision, and recall across multiple tasks, alongside marked reductions in GPU usage, training time, and memory consumption. The fine-tuned model also exhibited enhanced adaptability and generalization capabilities, performing well on new and unseen tasks, thus proving the efficacy and practical benefits of our methods. The contributions of this research provide valuable insights into optimizing large language models, broadening their applicability, and making advanced natural language processing techniques more accessible and efficient.\u003c/p\u003e","manuscriptTitle":"Unified Efficient Fine-Tuning Techniques for Open-Source Large Language Models","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-07-02 03:02:18","doi":"10.21203/rs.3.rs-4660140/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":"6bfa9c23-68d8-49f8-87be-1974dc8c528e","owner":[],"postedDate":"July 2nd, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":33899734,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2024-07-02T03:02:18+00:00","versionOfRecord":[],"versionCreatedAt":"2024-07-02 03:02:18","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4660140","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4660140","identity":"rs-4660140","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}