[{"data":1,"prerenderedAt":186},["ShallowReactive",2],{"content-query-011z4xTiLd":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"body":13,"_type":180,"_id":181,"_source":182,"_file":183,"_stem":184,"_extension":185},"/news/zh/2749","zh",false,"","百川智能 x 昇腾AI｜百川重磅发布Baichuan2大模型!已上线昇思开源社区","【2023年9月6日，北京】今日，百川智能在北京召开大模型发布会，正式发布Baichuan2开源大模型，昇腾AI基础软硬件平台正式支持Baichuan2大模型，并在昇思MindSpore开源社区大模型平台上线Baichuan2-7B模型开放体验。","2023-09-06","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2023/09/15/01034f77d30841aaaa6eb6eda16cbbd6.png","news",{"type":14,"children":15,"toc":177},"root",[16,24,29,34,43,48,59,64,74,100,105,112,117,122,129,134,141,146,151,159,164,172],{"type":17,"tag":18,"props":19,"children":21},"element","h1",{"id":20},"百川智能-x-昇腾ai百川重磅发布baichuan2大模型已上线昇思开源社区",[22],{"type":23,"value":8},"text",{"type":17,"tag":25,"props":26,"children":27},"p",{},[28],{"type":23,"value":9},{"type":17,"tag":25,"props":30,"children":31},{},[32],{"type":23,"value":33},"发布会现场，百川智能宣布正式开源微调后的Baichuan2-7B、Baichuan2-13B、Baichuan2-13B-Chat与其量化版本4bit，面向学术及商业市场提供大模型服务，并且均为免费可商用。",{"type":17,"tag":25,"props":35,"children":36},{},[37],{"type":17,"tag":38,"props":39,"children":42},"img",{"alt":40,"src":41},"image.png","https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20230913072017.28475439507823653709897863056339:50540913011914:2400:D41E90D54B08DE6A26075D1E984044D8024768488D536F6041CCFB28A20D5C33.png",[],{"type":17,"tag":25,"props":44,"children":45},{},[46],{"type":23,"value":47},"昇思MindSpore版本开源仓链接：",{"type":17,"tag":25,"props":49,"children":50},{},[51],{"type":17,"tag":52,"props":53,"children":57},"a",{"href":54,"rel":55},"https://gitee.com/mindspore/mindformers/blob/dev/research/baichuan2/baichuan2.md",[56],"nofollow",[58],{"type":23,"value":54},{"type":17,"tag":25,"props":60,"children":61},{},[62],{"type":23,"value":63},"昇思MindSpore开源社区大模型平台：",{"type":17,"tag":25,"props":65,"children":66},{},[67],{"type":17,"tag":52,"props":68,"children":71},{"href":69,"rel":70},"https://xihe.mindspore.cn/modelzoo/baichuan2%5C_7b%5C_chat",[56],[72],{"type":23,"value":73},"https://xihe.mindspore.cn/modelzoo/baichuan2\\_7b\\_chat",{"type":17,"tag":25,"props":75,"children":76},{},[77,91,93,98],{"type":17,"tag":78,"props":79,"children":80},"strong",{},[81],{"type":17,"tag":78,"props":82,"children":83},{},[84,89],{"type":17,"tag":78,"props":85,"children":86},{},[87],{"type":23,"value":88},"文理兼备性能优异，全",{"type":23,"value":90},"**面",{"type":23,"value":92},"领先** ",{"type":17,"tag":78,"props":94,"children":95},{},[96],{"type":23,"value":97},"LLaMA 2",{"type":23,"value":99},"****",{"type":17,"tag":25,"props":101,"children":102},{},[103],{"type":23,"value":104},"Baichuan2-7B-Base 和 Baichuan2-13B-Base，均基于2.6T⾼质量多语⾔数据进⾏训练，在保留了上一代开源模型良好的生成与创作能力，流畅的多轮对话能力以及部署⻔槛较低等众多特性的基础上，两个模型在数学、代码、安全、逻辑推理、语义理解等能⼒有显著提升。其中Baichuan2-13B-Base相比上⼀代13B模型，数学能力提升49%，代码能⼒提升46%，安全能力提升37%，逻辑推理能力提升25%，语义理解能力提升15%。",{"type":17,"tag":25,"props":106,"children":107},{},[108],{"type":17,"tag":38,"props":109,"children":111},{"alt":40,"src":110},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20230913072054.23023676250745622017721902109817:50540913011914:2400:FC20A8F7519CF01EEE473ECC7AC7A0AEF4EE69FBF080C6DB1D4639EC4D6FC384.png",[],{"type":17,"tag":25,"props":113,"children":114},{},[115],{"type":23,"value":116},"本次开源的两个模型在各⼤评测榜单上的表现优秀，在MMLU、CMMLU、GSM8K等几⼤权威评估基准中，以绝对优势领先LLaMA 2，相比其他同等参数量大模型，表现也十分亮眼，性能大幅度优于LLaMA 2等同尺寸模型竞品。",{"type":17,"tag":25,"props":118,"children":119},{},[120],{"type":23,"value":121},"更值得一提的是，根据MMLU等多个权威英文评估基准评分 Baichuan2-7b以70亿的参数在英文主流任务上与130 亿参数量的LLaMA 2基本持平。",{"type":17,"tag":25,"props":123,"children":124},{},[125],{"type":17,"tag":38,"props":126,"children":128},{"alt":40,"src":127},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20230913072112.85275182370566088263755807395289:50540913011914:2400:A0D0D85A3B9A6B8ED562A22E281FBBE400CBC3CD59E69A7F3BC7190BDE46ECD3.png",[],{"type":17,"tag":25,"props":130,"children":131},{},[132],{"type":23,"value":133},"7B参数模型的Benchmark成绩",{"type":17,"tag":25,"props":135,"children":136},{},[137],{"type":17,"tag":38,"props":138,"children":140},{"alt":40,"src":139},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20230913072133.37255981252003796254904711683763:50540913011914:2400:BD744B17F973C2318BB363E5D29CFFF1246178A0B7599D2A29DCB80DCA690D9E.png",[],{"type":17,"tag":25,"props":142,"children":143},{},[144],{"type":23,"value":145},"13B参数模型的Benchmark成绩",{"type":17,"tag":25,"props":147,"children":148},{},[149],{"type":23,"value":150},"Baichuan2-7B和Baichuan2-13B不仅对学术研究完全开放，开发者也仅需邮件申请获得官方商用许可后，即可以免费商用。",{"type":17,"tag":25,"props":152,"children":153},{},[154],{"type":17,"tag":78,"props":155,"children":156},{},[157],{"type":23,"value":158},"Baichuan2****大模型",{"type":17,"tag":25,"props":160,"children":161},{},[162],{"type":23,"value":163},"Baichuan2大模型是由百川智能开发的一系列开源可商用的大规模预训练语言模型。包含了70亿、130亿、530亿参数的模型。百川智能⾃成⽴之初，就将通过开源⽅式助⼒中国⼤模型⽣态繁荣作为公司的重要发展⽅向。此次开源的两款Baichuan2⼤模型，得到了上下游企业的积极响应，华为等众多知名企业均参加了本次发布会并与百川智能达成了合作。",{"type":17,"tag":25,"props":165,"children":166},{},[167],{"type":17,"tag":78,"props":168,"children":169},{},[170],{"type":23,"value":171},"昇腾AI",{"type":17,"tag":25,"props":173,"children":174},{},[175],{"type":23,"value":176},"昇腾 AI 是以昇腾 AI 基础软硬件平台为基础构建的人工智能计算产业。昇腾AI基础软硬件平台包括 Atlas 系列硬件及伙伴自有品牌硬件、异构计算架构 CANN、全场景AI框架昇思 MindSpore、昇腾应用使能 MindX、一站式开发平台 ModelArts 和统一工具链MindStudio 等。",{"title":7,"searchDepth":178,"depth":178,"links":179},4,[],"markdown","content:news:zh:2749.md","content","news/zh/2749.md","news/zh/2749","md",1776506072507]