[{"data":1,"prerenderedAt":344},["ShallowReactive",2],{"content-query-478idZP2Sx":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"body":13,"_type":338,"_id":339,"_source":340,"_file":341,"_stem":342,"_extension":343},"/news/zh/2861","zh",false,"","昆仑万维开源基于昇思MindSpore的「天工」13B系列大模型，0门槛可商用","10月30日，昆仑万维宣布开源百亿级大语言模型「天工」Skywork-13B系列，并罕见地配套开源了600GB、150B Tokens的超大高质量开源中文数据集。","2023-11-01","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2023/11/10/8882dd2049744cdeab6f9ac2425189cd.png","news",{"type":14,"children":15,"toc":335},"root",[16,24,30,57,70,75,84,89,94,102,110,115,141,146,186,191,211,216,221,229,234,243,248,256,261,269,274,281,289,294,302,307,312,317,325,330],{"type":17,"tag":18,"props":19,"children":21},"element","h1",{"id":20},"昆仑万维开源基于昇思mindspore的天工13b系列大模型0门槛可商用",[22],{"type":23,"value":8},"text",{"type":17,"tag":25,"props":26,"children":27},"p",{},[28],{"type":23,"value":29},"10月30日，昆仑万维宣布开源百亿级大语言模型「天工」Skywork-13B系列，并罕见地配套开源了600GB、150B Tokens的超大高质量开源中文数据集。该系列大模型利用昇思MindSpore AI框架完成模型开发，并将在昇思MindSpore开源社区上线大模型体验。",{"type":17,"tag":25,"props":31,"children":32},{},[33,35,41,43,48,50,55],{"type":23,"value":34},"昆仑万维「天工」Skywork-13B系列目前包括130亿参数的两大模型：",{"type":17,"tag":36,"props":37,"children":38},"strong",{},[39],{"type":23,"value":40},"Skywork-13B-Base模型、Skywork-13B-Math模型",{"type":23,"value":42},"，它们在CEVAL, GSM8K等多个权威评测与基准测试上都展现了同等规模模型的",{"type":17,"tag":36,"props":44,"children":45},{},[46],{"type":23,"value":47},"最佳效果",{"type":23,"value":49},"，其",{"type":17,"tag":36,"props":51,"children":52},{},[53],{"type":23,"value":54},"中文能力",{"type":23,"value":56},"尤为出色，在中文科技、金融、政务等领域表现均高于其他开源模型。",{"type":17,"tag":25,"props":58,"children":59},{},[60,62],{"type":23,"value":61},"Skywork-13B下载地址（Model Scope）：",{"type":17,"tag":63,"props":64,"children":68},"a",{"href":65,"rel":66},"https://modelscope.cn/organization/skywork",[67],"nofollow",[69],{"type":23,"value":65},{"type":17,"tag":25,"props":71,"children":72},{},[73],{"type":23,"value":74},"Skywork-13B下载地址（Github）：",{"type":17,"tag":25,"props":76,"children":77},{},[78],{"type":17,"tag":63,"props":79,"children":82},{"href":80,"rel":81},"https://github.com/SkyworkAI/Skywork",[67],[83],{"type":23,"value":80},{"type":17,"tag":25,"props":85,"children":86},{},[87],{"type":23,"value":88},"昇思MindSpore开源社区大模型平台：",{"type":17,"tag":25,"props":90,"children":91},{},[92],{"type":23,"value":93},"昇思大模型平台 (mindspore.cn)",{"type":17,"tag":25,"props":95,"children":96},{},[97],{"type":17,"tag":36,"props":98,"children":99},{},[100],{"type":23,"value":101},"除模型开源外，Skywork-13B系列大模型还将开源600GB、150B Tokens的高质量中文语料数据集Skypile/Chinese-Web-Text-150B，这是目前最大的开源中文数据集之一。",{"type":17,"tag":25,"props":103,"children":104},{},[105],{"type":17,"tag":36,"props":106,"children":107},{},[108],{"type":23,"value":109},"同时，昆仑万维「天工」Skywork-13B系列大模型即将全面开放商用——开发者无需申请，即可商用，堪称业内开源最彻底的百亿高质量商用模型。",{"type":17,"tag":25,"props":111,"children":112},{},[113],{"type":23,"value":114},"Skywork-13B系列大模型的开源将为大模型的场景应用和开源社区发展提供更佳的技术支持，降低大模型商业门槛，推动人工智能技术落地千行百业，为人工智能生态建设添砖加瓦，携手开源社区探索未知世界、创造美好未来。",{"type":17,"tag":25,"props":116,"children":117},{},[118],{"type":17,"tag":36,"props":119,"children":120},{},[121],{"type":17,"tag":36,"props":122,"children":123},{},[124],{"type":17,"tag":36,"props":125,"children":126},{},[127],{"type":17,"tag":36,"props":128,"children":129},{},[130],{"type":17,"tag":36,"props":131,"children":132},{},[133],{"type":17,"tag":36,"props":134,"children":135},{},[136],{"type":17,"tag":36,"props":137,"children":138},{},[139],{"type":23,"value":140},"两大模型 领先行业",{"type":17,"tag":25,"props":142,"children":143},{},[144],{"type":23,"value":145},"昆仑万维「天工」Skywork-13B系列包括两大模型及150B高质量中文数据集。",{"type":17,"tag":147,"props":148,"children":149},"ul",{},[150,161,171],{"type":17,"tag":151,"props":152,"children":153},"li",{},[154,159],{"type":17,"tag":36,"props":155,"children":156},{},[157],{"type":23,"value":158},"Skywork-13B-Base模型",{"type":23,"value":160},"是Skywork-13B的基础模型，其经由3.2万亿个多语言高质量数据训练，在CEVAL, CMMLU, MMLU, GSM8K等评测与基准测试上都展现了同等规模模型的最佳效果。",{"type":17,"tag":151,"props":162,"children":163},{},[164,169],{"type":17,"tag":36,"props":165,"children":166},{},[167],{"type":23,"value":168},"Skywork-13B-Math模型",{"type":23,"value":170},"经过专门的数学能力强化训练，在GSM8K等数据集上取得了同等规模模型的最佳效果。",{"type":17,"tag":151,"props":172,"children":173},{},[174,179,181],{"type":17,"tag":36,"props":175,"children":176},{},[177],{"type":23,"value":178},"Skypile/Chinese-Web-Text-150B数据集",{"type":23,"value":180},"是经过精心过滤的数据处理流程从中文网页中筛选出的高质量数据。",{"type":17,"tag":36,"props":182,"children":183},{},[184],{"type":23,"value":185},"本次开源的数据集大小约为600GB，总token数量约为150B，目前开源最大的中文数据集之一。",{"type":17,"tag":25,"props":187,"children":188},{},[189],{"type":23,"value":190},"除此之外，Skywork-13B系列还公开了模型中使用的评估方法、数据配比研究和训练基础设施调优方案等。希望这些开源内容能够进一步启发社区对于大型模型预训练的认知，并推动人工智能通用智能（AGI）的实现。",{"type":17,"tag":25,"props":192,"children":193},{},[194],{"type":17,"tag":36,"props":195,"children":196},{},[197],{"type":17,"tag":36,"props":198,"children":199},{},[200],{"type":17,"tag":36,"props":201,"children":202},{},[203],{"type":17,"tag":36,"props":204,"children":205},{},[206],{"type":17,"tag":36,"props":207,"children":208},{},[209],{"type":23,"value":210},"五大特点 全面超越",{"type":17,"tag":25,"props":212,"children":213},{},[214],{"type":23,"value":215},"昆仑万维「天工」Skywork-13B系列大模型在CEVAL, GSM8K等多个权威评测与基准测试上都展现了同等规模模型的最佳效果，其中文能力尤为出色，在中文科技、金融、政务等领域表现均高于其他开源模型。",{"type":17,"tag":25,"props":217,"children":218},{},[219],{"type":23,"value":220},"Skywork-13B系列模型的五大特点：",{"type":17,"tag":25,"props":222,"children":223},{},[224],{"type":17,"tag":36,"props":225,"children":226},{},[227],{"type":23,"value":228},"1、最强参数表现：全面超越同等规模大模型",{"type":17,"tag":25,"props":230,"children":231},{},[232],{"type":23,"value":233},"本次开源的Skywork-13B系列模型在CEVAL, CMMLU, MMLU, GSM8K等几大权威评估基准中全面超越LLaMA2-13B等开源大模型，在同等规模大模型间取得最佳效果。（截止至10月25日数据）",{"type":17,"tag":25,"props":235,"children":236},{},[237],{"type":17,"tag":238,"props":239,"children":242},"img",{"alt":240,"src":241},"image.png","https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20231110063048.73527509299267710731946225783485:50541109073030:2400:E1622E51879ABFF79ECD331493A2CB863987D48E9353D38853056DF3748A8EC6.png",[],{"type":17,"tag":25,"props":244,"children":245},{},[246],{"type":23,"value":247},"图2 基于时空演化解耦的三支路网络架构",{"type":17,"tag":25,"props":249,"children":250},{},[251],{"type":17,"tag":36,"props":252,"children":253},{},[254],{"type":23,"value":255},"2、最多训练数据：3.2T高质量的多语言训练数据",{"type":17,"tag":25,"props":257,"children":258},{},[259],{"type":23,"value":260},"Skywork-13B系列大模型拥有130亿参数、3.2万亿高质量多语言训练数据。在模型的生成能力、创作能力和数学推理能力提升明显。",{"type":17,"tag":25,"props":262,"children":263},{},[264],{"type":17,"tag":36,"props":265,"children":266},{},[267],{"type":23,"value":268},"3、最强中文语言建模能力：中文语言建模困惑度评测，表现优异",{"type":17,"tag":25,"props":270,"children":271},{},[272],{"type":23,"value":273},"Skywork-13B系列大模型在中文语言建模能力方面表现出色，具有很出色的中文文创能力。在中文文本创作领域的评测中，Skywork-13B系列大模型展现出了卓越的能力，尤其在科技、金融、政务、企业服务、文创、游戏等领域均表现高于业内其他开源模型。",{"type":17,"tag":25,"props":275,"children":276},{},[277],{"type":17,"tag":238,"props":278,"children":280},{"alt":240,"src":279},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20231110063109.13661486308383594006622555706389:50541109073030:2400:F16ACB1511380E26E8DDBB7870C8693088FEC1D4B1AC5F361D6B81E092F43F26.png",[],{"type":17,"tag":25,"props":282,"children":283},{},[284],{"type":17,"tag":36,"props":285,"children":286},{},[287],{"type":23,"value":288},"4、最大中文开源数据集之一：150B Tokens高质量中文语料",{"type":17,"tag":25,"props":290,"children":291},{},[292],{"type":23,"value":293},"Skywork-13B系列将配套开源600GB、150B Tokens的高质量中文语料数据集Skypile/Chinese-Web-Text-150B，这是目前最大的开源中文数据集之一。开发者可以最大程度地借鉴技术报告中大模型预训练的过程和经验，深度定制模型参数，有针对性的进行训练与优化。",{"type":17,"tag":25,"props":295,"children":296},{},[297],{"type":17,"tag":36,"props":298,"children":299},{},[300],{"type":23,"value":301},"5、诚意的开源商用：无需申请，即可实现商用",{"type":17,"tag":25,"props":303,"children":304},{},[305],{"type":23,"value":306},"目前开源社区中的中文大模型多数并非是完全可商用，一般开源社区用户通常需要进行复杂的商用授权申请流程，在某些情况，甚至有对公司规模、所在行业、用户数等维度有明确规定不给予商业授权。",{"type":17,"tag":25,"props":308,"children":309},{},[310],{"type":23,"value":311},"昆仑万维对Skywork-13B系列开源的开放性和可商用性高度重视，将授权流程做到极简，取消对行业、公司规模、用户等方面的限制，目的是帮助更多对中文大模型感兴趣的用户和企业在行业中不断探索和进步。",{"type":17,"tag":25,"props":313,"children":314},{},[315],{"type":23,"value":316},"此次Skywork-13B系列大模型将全面开放商用许可，用户在下载模型并同意并遵守《Skywork模型社区许可协议》后，无需再次申请授权即可将大模型进行商业用途。希望用户能够更便捷地探索Skywork-13B系列大模型技术能力，探索在不同场景下的商业化应用。",{"type":17,"tag":25,"props":318,"children":319},{},[320],{"type":17,"tag":36,"props":321,"children":322},{},[323],{"type":23,"value":324},"昆仑万维集团",{"type":17,"tag":25,"props":326,"children":327},{},[328],{"type":23,"value":329},"昆仑万维于2008年成立，2015年深交所上市，从游戏起家到AII In AGI与AIGC， 全面构建多元化的业务生态，至今十余年的发展，我们始终致力于为全球用户提供领先的互联网产品与服务。现今，昆仑万维还在不断探索AI领域的无限可能。目前昆仑万维逐渐构建了AGI与AIGC、海外信息分发与元宇宙、投资三大业务板块，业务覆盖全球一百多个国家和地区，全球平均月活跃用户近4亿。",{"type":17,"tag":25,"props":331,"children":332},{},[333],{"type":23,"value":334},"凭借对科技发展趋势的超前预判，昆仑万维早在2020年便已开始布局AIGC领域。至今，已积累近三年的相关工程研发经验，并建立了行业领先的预训练数据深度处理能力，昆仑万维也在人工智能领域取得了重大突破，目前已形成AI大模型、AI搜索、AI游戏、AI音乐、AI动漫、AI社交六大AI业务矩阵，是国内模型技术与工程能力最强，布局最全面，同时全身心投入开源社区建设的企业之一。",{"title":7,"searchDepth":336,"depth":336,"links":337},4,[],"markdown","content:news:zh:2861.md","content","news/zh/2861.md","news/zh/2861","md",1776506074557]