\n\t\t\n\t\n\t\n\t\tMobiLlama: Revolutionizing Efficient AI for Edge Devices\n\t\n\n

\n\n

\n\t\n\t\t\n\t\n\t\n\t\tLinks 🔗:\n\t\n

👉 Subscribe: https://www.youtube.com/@Arxflix
👉 Twitter: https://x.com/arxflix
👉 LMNT (Partner): https://lmnt.com/

\n","updatedAt":"2024-06-09T05:06:29.108Z","author":{"_id":"6186ddf6a7717cb375090c01","avatarUrl":"/avatars/716b6a7d1094c8036b2a8a7b9063e8aa.svg","fullname":"Julien BLANCHON","name":"blanchon","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":176,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.48521843552589417},"editors":["blanchon"],"editorAvatarUrls":["/avatars/716b6a7d1094c8036b2a8a7b9063e8aa.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2402.16840","authors":[{"_id":"65dd76e84a4fce1ec96347f4","user":{"_id":"64b7d2ad8c632fbca9507431","avatarUrl":"/avatars/76c31ea218108cf6c3715269f7605404.svg","isPro":false,"fullname":"Omkar Thawakar","user":"omkarthawakar","type":"user"},"name":"Omkar Thawakar","status":"admin_assigned","statusLastChangedAt":"2024-02-27T10:29:58.983Z","hidden":false},{"_id":"65dd76e84a4fce1ec96347f5","user":{"_id":"64f9d0ded88459d40bd0efb9","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/64f9d0ded88459d40bd0efb9/16Y41IUX0T_TZj_g_tTBF.jpeg","isPro":false,"fullname":"Ashmal Vayani","user":"Ashmal","type":"user"},"name":"Ashmal Vayani","status":"admin_assigned","statusLastChangedAt":"2024-02-27T10:30:14.795Z","hidden":false},{"_id":"65dd76e84a4fce1ec96347f6","name":"Salman Khan","hidden":false},{"_id":"65dd76e84a4fce1ec96347f7","user":{"_id":"654a5f4f9b8bd6406d45bb46","avatarUrl":"/avatars/ac0d7eef62cd98a280b162cf7896b1a2.svg","isPro":false,"fullname":"Hisham Cholakkal","user":"hishamcholakkal","type":"user"},"name":"Hisham Cholakal","status":"admin_assigned","statusLastChangedAt":"2024-02-27T10:30:47.610Z","hidden":false},{"_id":"65dd76e84a4fce1ec96347f8","name":"Rao M. Anwer","hidden":false},{"_id":"65dd76e84a4fce1ec96347f9","user":{"_id":"63d38a1a9e1b81818c816905","avatarUrl":"/avatars/a2160bb9fbb8a8eec129b8eab559294b.svg","isPro":false,"fullname":"Michael Felsberg","user":"CVL-LiU","type":"user"},"name":"Michael Felsberg","status":"admin_assigned","statusLastChangedAt":"2024-02-27T10:31:08.550Z","hidden":false},{"_id":"65dd76e84a4fce1ec96347fa","user":{"_id":"65d895d8d435dce78392065f","avatarUrl":"/avatars/68c7185d3dd1adf78927b91849978d62.svg","isPro":false,"fullname":"Tim Baldwin","user":"tbaldwin","type":"user"},"name":"Tim Baldwin","status":"admin_assigned","statusLastChangedAt":"2024-02-27T10:31:14.438Z","hidden":false},{"_id":"65dd76e84a4fce1ec96347fb","user":{"_id":"64ff67722ad36636be6c4542","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/sLIrNelAWPVOy4e3oo5LB.jpeg","isPro":false,"fullname":"Eric Xing","user":"EricX003","type":"user"},"name":"Eric P. Xing","status":"admin_assigned","statusLastChangedAt":"2024-02-27T10:31:25.777Z","hidden":false},{"_id":"65dd76e84a4fce1ec96347fc","name":"Fahad Shahbaz Khan","hidden":false}],"publishedAt":"2024-02-26T18:59:03.000Z","submittedOnDailyAt":"2024-02-27T03:15:13.066Z","title":"MobiLlama: Towards Accurate and Lightweight Fully Transparent GPT","submittedOnDailyBy":{"_id":"60f1abe7544c2adfd699860c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674929746905-60f1abe7544c2adfd699860c.jpeg","isPro":false,"fullname":"AK","user":"akhaliq","type":"user"},"summary":"\"Bigger the better\" has been the predominant trend in recent Large Language\nModels (LLMs) development. However, LLMs do not suit well for scenarios that\nrequire on-device processing, energy efficiency, low memory footprint, and\nresponse efficiency. These requisites are crucial for privacy, security, and\nsustainable deployment. This paper explores the \"less is more\" paradigm by\naddressing the challenge of designing accurate yet efficient Small Language\nModels (SLMs) for resource constrained devices. Our primary contribution is the\nintroduction of an accurate and fully transparent open-source 0.5 billion\n(0.5B) parameter SLM, named MobiLlama, catering to the specific needs of\nresource-constrained computing with an emphasis on enhanced performance with\nreduced resource demands. MobiLlama is a SLM design that initiates from a\nlarger model and applies a careful parameter sharing scheme to reduce both the\npre-training and the deployment cost. Our work strives to not only bridge the\ngap in open-source SLMs but also ensures full transparency, where complete\ntraining data pipeline, training code, model weights, and over 300 checkpoints\nalong with evaluation codes is available at :\nhttps://github.com/mbzuai-oryx/MobiLlama.","upvotes":25,"discussionId":"65dd76e94a4fce1ec9634815","githubRepo":"https://github.com/mbzuai-oryx/mobillama","githubRepoAddedBy":"auto","ai_summary":"MobiLlama is a 0.5 billion parameter small language model designed for resource-constrained devices using parameter sharing to balance accuracy and efficiency.","ai_keywords":["Large Language Models","Small Language Models","resource constrained devices","parameter sharing","MobiLlama","pre-training","deployment cost","full transparency","training data pipeline","training code","model weights","checkpoints","evaluation codes"],"githubStars":668},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"620783f24e28382272337ba4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/620783f24e28382272337ba4/zkUveQPNiDfYjgGhuFErj.jpeg","isPro":false,"fullname":"GuoLiangTang","user":"Tommy930","type":"user"},{"_id":"62d1ddfac58f969c1528f1b5","avatarUrl":"/avatars/75c372a831cde3c7c6dce3bc875488a7.svg","isPro":false,"fullname":"Kalle Hilsenbek","user":"Bachstelze","type":"user"},{"_id":"6362ddb7d3be91534c30bfd6","avatarUrl":"/avatars/dac76ebd3b8a08099497ec0b0524bc7c.svg","isPro":false,"fullname":"Art Atk","user":"ArtAtk","type":"user"},{"_id":"6538119803519fddb4a17e10","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6538119803519fddb4a17e10/ffJMkdx-rM7VvLTCM6ri_.jpeg","isPro":false,"fullname":"samusenps","user":"samusenps","type":"user"},{"_id":"64747f7e33192631bacd8831","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/64747f7e33192631bacd8831/dstkZJ4sHJSeqLesV5cOC.jpeg","isPro":false,"fullname":"Taufiq Dwi Purnomo","user":"taufiqdp","type":"user"},{"_id":"6396e5e14933295d5c9ae306","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1670833600406-noauth.png","isPro":false,"fullname":"KeronDev","user":"Keron","type":"user"},{"_id":"64b7d2ad8c632fbca9507431","avatarUrl":"/avatars/76c31ea218108cf6c3715269f7605404.svg","isPro":false,"fullname":"Omkar Thawakar","user":"omkarthawakar","type":"user"},{"_id":"6378fcde667ed02bb915cdc3","avatarUrl":"/avatars/69a7abaee92f4e53bf722f8a0833b2b1.svg","isPro":false,"fullname":"Vaibhav Singh","user":"veb-101","type":"user"},{"_id":"641a6c5919fc5647be1918f8","avatarUrl":"/avatars/c03297c68f90b9cdb7adc1f21c724dce.svg","isPro":false,"fullname":"David Euler","user":"davideuler","type":"user"},{"_id":"6039478ab3ecf716b1a5fd4d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg","isPro":true,"fullname":"taesiri","user":"taesiri","type":"user"},{"_id":"65953f3a078efa3255dae953","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/26H96hHlCmCf8a31TKV48.jpeg","isPro":false,"fullname":"Mosaab Muhammad","user":"Mosaabx","type":"user"},{"_id":"654a5f4f9b8bd6406d45bb46","avatarUrl":"/avatars/ac0d7eef62cd98a280b162cf7896b1a2.svg","isPro":false,"fullname":"Hisham Cholakkal","user":"hishamcholakkal","type":"user"}],"acceptLanguages":["*"],"dailyPaperRank":0}">

Papers

arxiv:2402.16840

MobiLlama: Towards Accurate and Lightweight Fully Transparent GPT

Published on Feb 26, 2024

· Submitted by

AK on Feb 27, 2024

Upvote

Authors:

Omkar Thawakar ,

Ashmal Vayani ,

Hisham Cholakal ,

Michael Felsberg ,

Tim Baldwin ,

Eric P. Xing ,

Abstract

MobiLlama is a 0.5 billion parameter small language model designed for resource-constrained devices using parameter sharing to balance accuracy and efficiency.

AI-generated summary

"Bigger the better" has been the predominant trend in recent Large Language Models (LLMs) development. However, LLMs do not suit well for scenarios that require on-device processing, energy efficiency, low memory footprint, and response efficiency. These requisites are crucial for privacy, security, and sustainable deployment. This paper explores the "less is more" paradigm by addressing the challenge of designing accurate yet efficient Small Language Models (SLMs) for resource constrained devices. Our primary contribution is the introduction of an accurate and fully transparent open-source 0.5 billion (0.5B) parameter SLM, named MobiLlama, catering to the specific needs of resource-constrained computing with an emphasis on enhanced performance with reduced resource demands. MobiLlama is a SLM design that initiates from a larger model and applies a careful parameter sharing scheme to reduce both the pre-training and the deployment cost. Our work strives to not only bridge the gap in open-source SLMs but also ensures full transparency, where complete training data pipeline, training code, model weights, and over 300 checkpoints along with evaluation codes is available at : https://github.com/mbzuai-oryx/MobiLlama.