[{"data":1,"prerenderedAt":1169},["ShallowReactive",2],{"content-query-T6c2tUoxQi":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"body":13,"_type":1163,"_id":1164,"_source":1165,"_file":1166,"_stem":1167,"_extension":1168},"/news/zh/3678","zh",false,"","社区季报 | 新版本发布、开发者相约，MindSpore Developer Day 2025重磅来袭！","昇思MindSpore2025年一季度总结","2025-04-09","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/5a5811fb92fd45bebd0266169ce74ab1.png","news",{"type":14,"children":15,"toc":1160},"root",[16,24,37,45,50,55,60,65,70,75,83,88,93,112,120,128,136,141,146,153,161,166,171,182,187,196,203,211,216,223,231,236,243,251,256,261,268,276,284,289,300,307,315,320,327,335,340,347,355,360,367,375,380,387,395,400,407,415,420,427,435,440,448,453,460,468,473,480,488,493,500,508,513,520,528,533,540,548,553,560,568,573,580,588,593,600,608,613,620,628,633,640,648,653,658,667,676,683,691,699,704,709,718,727,734,742,747,752,759,767,772,784,791,799,804,811,819,824,829,839,846,854,859,864,871,879,884,889,893,902,907,916,923,931,936,941,952,959,967,972,977,988,999,1006,1014,1019,1024,1031,1039,1044,1051,1059,1064,1071,1079,1084,1089,1094,1101,1109,1114,1119,1130,1137,1150,1155],{"type":17,"tag":18,"props":19,"children":21},"element","h1",{"id":20},"社区季报-新版本发布开发者相约mindspore-developer-day-2025重磅来袭",[22],{"type":23,"value":8},"text",{"type":17,"tag":25,"props":26,"children":27},"p",{},[28],{"type":17,"tag":29,"props":30,"children":31},"strong",{},[32],{"type":17,"tag":33,"props":34,"children":35},"em",{},[36],{"type":23,"value":9},{"type":17,"tag":25,"props":38,"children":39},{},[40],{"type":17,"tag":29,"props":41,"children":42},{},[43],{"type":23,"value":44},"一、概述",{"type":17,"tag":25,"props":46,"children":47},{},[48],{"type":23,"value":49},"2025年第一季度，昇思MindSpore开源社区以技术创新为驱动，持续扩大生态影响力，在框架升级、开发者生态建设、行业应用落地等方面取得显著进展。",{"type":17,"tag":25,"props":51,"children":52},{},[53],{"type":23,"value":54},"昇思MindSpore开源社区将于4月12日在杭州举办MindSpore Developer Day 2025。本次会议将发布MindSpore 2.6.0全新版本，展示其端到端全流程支持DeepSeek等MoE大模型训练、微调、强化学习与推理部署的技术，以及在科学智能、开发易用性、与CANN协同提升计算效率等领域和方向的技术创新与成果。同时联合模力方舟（GiteeAI）举办AI应用开发者Meetup，分享各行业领域AI应用开发实践经验与开发心得。",{"type":17,"tag":25,"props":56,"children":57},{},[58],{"type":23,"value":59},"在社区合作方面，社区与法本信息、三象智能、木愚科技、魔泊云、奥看科技、珑京科技、倍特威视、星舟志屹、易库智能等10+企业合作，基于DeepSeek、QwQ-32B等大模型打造了智能工作流、教育一体机、工业检测等解决方案，覆盖金融、教育、医疗、能源等多个行业，加速企业数智化转型。",{"type":17,"tag":25,"props":61,"children":62},{},[63],{"type":23,"value":64},"在技术方面，目前昇思MindSpore支持多项前沿技术与开源模型，包括支持文生视频SOTA模型（如Step-Video-T2V、Wan2.1）、强化学习框架GRPO-Training-Suite，以及原生适配QwQ-32B等大模型，显著提升了开发效率与性能表现。",{"type":17,"tag":25,"props":66,"children":67},{},[68],{"type":23,"value":69},"在社区运作方面，社区通过举办“昇思五日谈”系列技术直播、开源实习计划、量子计算黑客松大赛等活动，吸引了全球开发者深度参与，持续聚焦AI根技术研发，强化大模型与科学智能能力，为产业智能化升级提供坚实支撑。",{"type":17,"tag":25,"props":71,"children":72},{},[73],{"type":23,"value":74},"以下是2025年第一季度昇思MindSpore社区进展的详细报告。",{"type":17,"tag":25,"props":76,"children":77},{},[78],{"type":17,"tag":29,"props":79,"children":80},{},[81],{"type":23,"value":82},"二、社区规模",{"type":17,"tag":25,"props":84,"children":85},{},[86],{"type":23,"value":87},"截至3月底，昇思MindSpore面向全球开发者，凝聚产业力量，持续打造人工智能创新之源。社区秉持开放透明原则，稳步建设开源生态，规模和活跃度得到持续提升。",{"type":17,"tag":25,"props":89,"children":90},{},[91],{"type":23,"value":92},"社区发展了**4.5万+**核心贡献者；",{"type":17,"tag":25,"props":94,"children":95},{},[96,98,103,105,110],{"type":23,"value":97},"累计产生ISSUE共",{"type":17,"tag":29,"props":99,"children":100},{},[101],{"type":23,"value":102},"41.4****K",{"type":23,"value":104},"、PR共",{"type":17,"tag":29,"props":106,"children":107},{},[108],{"type":23,"value":109},"109K",{"type":23,"value":111},"。",{"type":17,"tag":25,"props":113,"children":114},{},[115],{"type":17,"tag":116,"props":117,"children":119},"img",{"alt":7,"src":118},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/6a907ca02e9d41028c10bb2047a0d30d.png",[],{"type":17,"tag":25,"props":121,"children":122},{},[123],{"type":17,"tag":29,"props":124,"children":125},{},[126],{"type":23,"value":127},"三、社区大事件",{"type":17,"tag":25,"props":129,"children":130},{},[131],{"type":17,"tag":29,"props":132,"children":133},{},[134],{"type":23,"value":135},"1、你有一封MindSpore Developer Day 2025邀请函，请查收",{"type":17,"tag":25,"props":137,"children":138},{},[139],{"type":23,"value":140},"为加速人工智能技术创新与产业落地，共建开源开放的AI框架技术生态昇思MindSpore开源社区将于2025年4月12日在杭州萧山万怡酒店举办MindSpore Developer Day 2025。",{"type":17,"tag":25,"props":142,"children":143},{},[144],{"type":23,"value":145},"本次会议聚焦AI框架技术创新与开源开发者生态共建，将发布MindSpore 2.6.0全新版本，展示其端到端全流程支持DeepSeek等MoE大模型训练、微调、强化学习与推理部署的技术，以及在科学智能、开发易用性、与CANN协同提升计算效率等领域和方向的技术创新与成果。同时昇思社区联合模力方舟（GiteeAI）举办AI应用开发者Meetup，分享各行业领域AI应用开发实践经验与开发心得，为大家带来产研前线的第一手干货。",{"type":17,"tag":25,"props":147,"children":148},{},[149],{"type":17,"tag":116,"props":150,"children":152},{"alt":7,"src":151},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/adce529181224d7eba9d6975950292be.png",[],{"type":17,"tag":25,"props":154,"children":155},{},[156],{"type":17,"tag":29,"props":157,"children":158},{},[159],{"type":23,"value":160},"2、训推全面支持、开箱即用！DeepSeek V3昇思MindSpore版本上线开源社区",{"type":17,"tag":25,"props":162,"children":163},{},[164],{"type":23,"value":165},"基于昇腾AI硬件与昇思MindSpore AI框架版本的DeepSeek-V3完成开发支持并上线昇思开源社区，面向开发者提供开箱即用的预训练和推理能力，并已成功在大规模集群上预训练和部署。通过获取昇思MindSpore版DeepSeek V3开源镜像，开发者可直接进行DeepSeek-V3的预训练和推理部署。",{"type":17,"tag":25,"props":167,"children":168},{},[169],{"type":23,"value":170},"昇思MindSpore开源社区训练代码：",{"type":17,"tag":25,"props":172,"children":173},{},[174],{"type":17,"tag":175,"props":176,"children":180},"a",{"href":177,"rel":178},"https://gitee.com/mindspore/mindformers/tree/dev/research/deepseek3",[179],"nofollow",[181],{"type":23,"value":177},{"type":17,"tag":25,"props":183,"children":184},{},[185],{"type":23,"value":186},"魔乐社区推理代码：",{"type":17,"tag":25,"props":188,"children":189},{},[190],{"type":17,"tag":175,"props":191,"children":194},{"href":192,"rel":193},"https://modelers.cn/models/MindSpore-Lab/DeepSeek-V3",[179],[195],{"type":23,"value":192},{"type":17,"tag":25,"props":197,"children":198},{},[199],{"type":17,"tag":116,"props":200,"children":202},{"alt":7,"src":201},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/ae2c27c6e2944e4786af09273c512bc0.png",[],{"type":17,"tag":25,"props":204,"children":205},{},[206],{"type":17,"tag":29,"props":207,"children":208},{},[209],{"type":23,"value":210},"3、携手北京大学，MindSpore+openEuler打造支持vLLM的DeepSeek全栈开源推理方案",{"type":17,"tag":25,"props":212,"children":213},{},[214],{"type":23,"value":215},"北京大学联合OpenAtom openEuler（简称\"openEuler\") 开源社区与MindSpore社区，推出面向大模型的全栈开源方案，以操作系统+AI框架+模型生态的三层开放架构，替换操作系统和DL框架，秉承代码开源+标准开放+生态共建的理念，逐步成为智能时代的自主创新数字基座。",{"type":17,"tag":25,"props":217,"children":218},{},[219],{"type":17,"tag":116,"props":220,"children":222},{"alt":7,"src":221},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/c0a4b44254324942953d61819ce442bc.png",[],{"type":17,"tag":25,"props":224,"children":225},{},[226],{"type":17,"tag":29,"props":227,"children":228},{},[229],{"type":23,"value":230},"4、昇思MindSpore亮相2025年全球开发者先锋大会（2025 GDC）",{"type":17,"tag":25,"props":232,"children":233},{},[234],{"type":23,"value":235},"2025年全球开发者先锋大会（2025 GDC）于2025年2月21日-23日在上海举办。会议期间，昇思MindSpore技术专家林翰在“融”见AI新生态，驱动经济新发展讲坛发表“大模型时代首选AI框架-昇思MindSpore”主题演讲。当前，昇思AI框架赋能用户开发大模型能力，充分发挥了硬件澎湃算力，同时兼顾开发效率与执行性能，架构、套件也在持续升级。",{"type":17,"tag":25,"props":237,"children":238},{},[239],{"type":17,"tag":116,"props":240,"children":242},{"alt":7,"src":241},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/23265f6d03604d469d291f7a08d6671b.png",[],{"type":17,"tag":25,"props":244,"children":245},{},[246],{"type":17,"tag":29,"props":247,"children":248},{},[249],{"type":23,"value":250},"5、MindSpore Quantum 0.10.0 正式发布，量子算法开发更灵活、更高效",{"type":17,"tag":25,"props":252,"children":253},{},[254],{"type":23,"value":255},"MindSpore Quantum是昇思MindSpore社区推出的全新量子经典混合计算框架，能够高效地支持主流的量子算法和经典量子混合算法，不但在量子机器学习、量子化学模拟和量子组合优化等变分量子算法上有着卓越的性能，而且在量子化学模拟规模上也处于领先地位，为量子计算的研究和发展提供了一个高效的开发环境。",{"type":17,"tag":25,"props":257,"children":258},{},[259],{"type":23,"value":260},"经过开发者们几个月的精心打磨，MindSpore Quantum 0.10.0 版本正式上线！本次版本不仅升级了量子模拟器和线路编译工具，更引入了全新的高维量子系统支持，助力科研创新和实际应用。",{"type":17,"tag":25,"props":262,"children":263},{},[264],{"type":17,"tag":116,"props":265,"children":267},{"alt":7,"src":266},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/126a7060b6b247aebd550af94ff39ff6.png",[],{"type":17,"tag":25,"props":269,"children":270},{},[271],{"type":17,"tag":29,"props":272,"children":273},{},[274],{"type":23,"value":275},"四、社区动态",{"type":17,"tag":25,"props":277,"children":278},{},[279],{"type":17,"tag":29,"props":280,"children":281},{},[282],{"type":23,"value":283},"1、昇思五日谈 | 5场直播，深度解析昇思支持DeepSeek训推全流程核心技术",{"type":17,"tag":25,"props":285,"children":286},{},[287],{"type":23,"value":288},"昇思MindSpore开源社区即将迎来开源五周年，为回馈开发者长期以来的支持与热爱，昇思MindSpore以一场硬核技术盛宴——「昇思五日谈」专题直播重磅上线！5场直播揭秘昇思MindSpore大模型训练与推理关键技术，助力开发者抢占AI技术高地！",{"type":17,"tag":25,"props":290,"children":291},{},[292,294],{"type":23,"value":293},"直播回放链接：",{"type":17,"tag":175,"props":295,"children":298},{"href":296,"rel":297},"https://space.bilibili.com/526894060/lists/4660795?type=series",[179],[299],{"type":23,"value":296},{"type":17,"tag":25,"props":301,"children":302},{},[303],{"type":17,"tag":116,"props":304,"children":306},{"alt":7,"src":305},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/3e16180033884d689c14708728d66d58.png",[],{"type":17,"tag":25,"props":308,"children":309},{},[310],{"type":17,"tag":29,"props":311,"children":312},{},[313],{"type":23,"value":314},"2、2025·昇思MindSpore开源实习启动",{"type":17,"tag":25,"props":316,"children":317},{},[318],{"type":23,"value":319},"2025·昇思MindSpore开源实习已经启动，欢迎各位高校学子报名参加。实习方法：1）线上实习，不论在家还是宿舍都可参与实习，积累项目开发经验，为实习简历添砖加瓦；2）实习薪资丰厚，完成任务获得积分即可在次月底发放对应薪资；3）任务种类繁多，从易到难各个程度的任务均有，总有你的专属项。2025·昇思MindSpore开源实习等你来“打卡”。",{"type":17,"tag":25,"props":321,"children":322},{},[323],{"type":17,"tag":116,"props":324,"children":326},{"alt":7,"src":325},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/c65fd3c9673b4c44938fc061c3a9a209.png",[],{"type":17,"tag":25,"props":328,"children":329},{},[330],{"type":17,"tag":29,"props":331,"children":332},{},[333],{"type":23,"value":334},"3、【实战演练】开发基于昇思MindSpore Demo，昇思Demo制作人C位出道吧！",{"type":17,"tag":25,"props":336,"children":337},{},[338],{"type":23,"value":339},"2月底，昇思MindSpore发布在线征集基于昇思MindSpore学习型项目案例的活动！无论是基于MindSpore的DeepSeek项目，还是开发「AI健身教练」「智能垃圾分类」等接地气项目，只要技术够硬，均可参加！此外，昇思MindSpore邀请开发者参与模型迁移挑战，将指定模型基于MindSpore NLP套件进行模型权重迁移，并上传至魔乐社区和大模型平台，均可获得相应激励！",{"type":17,"tag":25,"props":341,"children":342},{},[343],{"type":17,"tag":116,"props":344,"children":346},{"alt":7,"src":345},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/2412039da82948deaf9a522eef86a548.png",[],{"type":17,"tag":25,"props":348,"children":349},{},[350],{"type":17,"tag":29,"props":351,"children":352},{},[353],{"type":23,"value":354},"4、第七届·2025 MindSpore量子计算黑客松全国大赛启动报名",{"type":17,"tag":25,"props":356,"children":357},{},[358],{"type":23,"value":359},"一年一度的黑客松品牌赛事，“第七届·2025 MindSpore量子计算黑客松全国大赛”火热来袭！热身赛、量子组合优化、量子误差缓解、量子启发算法赛道目前已发布赛题，开发者将体验全新量子计算框架 MindSpore Quantum，本次赛题也将用到量子真机数据。诚邀各位开发者报名参赛，勇攀量子世界的高峰，挑战黑客松年度总冠军!",{"type":17,"tag":25,"props":361,"children":362},{},[363],{"type":17,"tag":116,"props":364,"children":366},{"alt":7,"src":365},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/95e92c2b493447e58e3d2db900cca66d.png",[],{"type":17,"tag":25,"props":368,"children":369},{},[370],{"type":17,"tag":29,"props":371,"children":372},{},[373],{"type":23,"value":374},"5、基于昇思实现古画修复，「昇思文旅SIG创新技术学习交流会」在西安成功举办",{"type":17,"tag":25,"props":376,"children":377},{},[378],{"type":23,"value":379},"3月29日， MindSpore Cultural Tourism SIG（文旅SIG）与民歌智能计算与服务技术文旅部重点实验室在陕西师范大学雁塔校区举办《昇思文旅SIG创新技术学习交流会》。本次会议围绕昇思Mindspore和文旅主题，面向所有对AI文旅感兴趣的开发者，希望通过学习昇思MindSpore的最新技术路线与基于昇思的DeepSeek实践，学习AI文旅的有趣案例与实践，共同孵化基于昇思的文旅创新应用。",{"type":17,"tag":25,"props":381,"children":382},{},[383],{"type":17,"tag":116,"props":384,"children":386},{"alt":7,"src":385},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/900f2e47478847fca95a0507e16046f0.png",[],{"type":17,"tag":25,"props":388,"children":389},{},[390],{"type":17,"tag":29,"props":391,"children":392},{},[393],{"type":23,"value":394},"6、中国物理学会-量子科技长三角产业创新中心-MindSpore Quantum学术奖励基金（二期）申请指南",{"type":17,"tag":25,"props":396,"children":397},{},[398],{"type":23,"value":399},"《中国物理学会-量子科技长三角产业创新中心-MindSpore Quantum学术奖励基金》是由中国物理学会、量子科技长三角产业创新中心和昇思MindSpore共同发起，致力于面向高校及科研院所的量子计算学者搭建学术交流平台，提供资金和技术支持、促进合作，推动开源量子计算软件在量子信息领域的广泛应用，并支持基于开源量子计算软件框架的国内外高水平期刊和会议的学术论文发表，激励原创性科学研究开展，为未来量子计算产业储备人才，构建中国量子计算研究的全球影响力。高校/科研院所的在职教师或研究人员（包括博士后）和在读博士研究生欢迎申报。",{"type":17,"tag":25,"props":401,"children":402},{},[403],{"type":17,"tag":116,"props":404,"children":406},{"alt":7,"src":405},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/d3ae20c7441944809868e4ad988ef048.png",[],{"type":17,"tag":25,"props":408,"children":409},{},[410],{"type":17,"tag":29,"props":411,"children":412},{},[413],{"type":23,"value":414},"7、「她力量·代码之外」：当技术温度遇见跨界创造力",{"type":17,"tag":25,"props":416,"children":417},{},[418],{"type":23,"value":419},"在三八妇女节之际，昇思MindSpore开源社区邀请了三位优秀女性开发者——中国科学院大学博士生孙宛奇、北京大学研究生王洁怡、瑞士巴塞尔大学神经生物学博士呼波，讲述她们以跨界思维、社区联结力与技术向善的初心，在量子计算、生物医学、自然语言等领域中绽放光彩的经历。她们的故事，既是科技女性的成长缩影，也是开源精神的最佳诠释。",{"type":17,"tag":25,"props":421,"children":422},{},[423],{"type":17,"tag":116,"props":424,"children":426},{"alt":7,"src":425},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/ea76ad3e1ccb444d9d8435e2c99bdf41.png",[],{"type":17,"tag":25,"props":428,"children":429},{},[430],{"type":17,"tag":29,"props":431,"children":432},{},[433],{"type":23,"value":434},"五、社区案例",{"type":17,"tag":25,"props":436,"children":437},{},[438],{"type":23,"value":439},"昇思MindSpore开源社区始终秉持“开放协作、技术赋能”的理念，积极推动大模型技术创新与行业实践深度融合，持续为开发者和企业提供前沿技术支撑与优化能力，助力大模型训练、微调、推理及场景应用创新。在2025年第一季度，社区联合生态伙伴，在金融、教育、医疗、能源、工业等多个行业孵化出一批解决方案，以下为部分案例的详细报告：",{"type":17,"tag":25,"props":441,"children":442},{},[443],{"type":17,"tag":29,"props":444,"children":445},{},[446],{"type":23,"value":447},"1、法本信息携手昇思：基于DeepSeek&QwQ-32B等打造企业级AI解决方案新范式",{"type":17,"tag":25,"props":449,"children":450},{},[451],{"type":23,"value":452},"在数字经济与实体经济深度融合的浪潮下，法本信息作为国内软件与信息技术服务领域的领军企业，基于昇腾AI基础软硬件平台及昇思MindSpore开源框架，构建“AI大模型+行业场景”的创新解决方案。依托昇腾强大的算力底座、高效的昇思AI框架、DeepSeek&QwQ-32B等国内知名大模型，以及法本信息在垂直领域的场景化落地能力，三方联合打造了覆盖智能编码、知识管理、智能制造、安全生产等领域的全栈式AI解决方案，为企业智能化转型提供技术引擎。",{"type":17,"tag":25,"props":454,"children":455},{},[456],{"type":17,"tag":116,"props":457,"children":459},{"alt":7,"src":458},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/834580672d904a6f91057dc993c3a182.png",[],{"type":17,"tag":25,"props":461,"children":462},{},[463],{"type":17,"tag":29,"props":464,"children":465},{},[466],{"type":23,"value":467},"2、三象智能基于昇思MindSpore，以DeepSeek&QwQ-32B等为基础大模型打造智能工作流升级解决方案",{"type":17,"tag":25,"props":469,"children":470},{},[471],{"type":23,"value":472},"在数字化转型加速的今天，企业如何高效处理海量对话数据？如何确保文档对比的精准性与时效性？如何以有限人力应对广告内容审核的复杂需求？三象智能基于昇思MindSpore，为企业打造智能工作流升级解决方案，依托DeepSeek R1&QwQ-32B等大模型的强大能力，从精准对话到智能风控，每个功能升级都为数字化转型注入新动能。",{"type":17,"tag":25,"props":474,"children":475},{},[476],{"type":17,"tag":116,"props":477,"children":479},{"alt":7,"src":478},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/cdc18deb5fa946fb88bec7bd22635f73.png",[],{"type":17,"tag":25,"props":481,"children":482},{},[483],{"type":17,"tag":29,"props":484,"children":485},{},[486],{"type":23,"value":487},"3、木愚科技携手昇思，以千问QwQ32B&DeepSeek等大模型为基础推出全栈式智能教育解决方案",{"type":17,"tag":25,"props":489,"children":490},{},[491],{"type":23,"value":492},"3月20日，深圳华为中国合作伙伴大会期间，木愚科技基于昇思与QwQ32B、DeepSeek等多个AI大模型，发布新一代“AI教育一体机”，以“AI+教育”深度融合为核心，为学校场景打造全栈式智能教育解决方案，助力教育提质增效与AI人才培养的双重目标。",{"type":17,"tag":25,"props":494,"children":495},{},[496],{"type":17,"tag":116,"props":497,"children":499},{"alt":7,"src":498},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/f6d212edfc3e44f0b630110320c15bef.png",[],{"type":17,"tag":25,"props":501,"children":502},{},[503],{"type":17,"tag":29,"props":504,"children":505},{},[506],{"type":23,"value":507},"4、西安电子科技大学人工智能学院携手昇思以千问QwQ-32B&DeepSeek等大模型为抓手，探索智慧教育，智慧科研新方式",{"type":17,"tag":25,"props":509,"children":510},{},[511],{"type":23,"value":512},"西安电子科技大学作为“人工智能+教育”标杆大学，探索出了基于昇思与QwQ-32B、DeepSeek等多个AI大模型的新一代\"AI教育科研协同平台\"，该平台支持万亿参数模型的本地化训练与实时推理，响应速度提升40%，能耗降低35%，满足高校对高效运算与节约能耗的双重诉求。",{"type":17,"tag":25,"props":514,"children":515},{},[516],{"type":17,"tag":116,"props":517,"children":519},{"alt":7,"src":518},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/3ec37208066f496ca12e0965a3e89a5d.png",[],{"type":17,"tag":25,"props":521,"children":522},{},[523],{"type":17,"tag":29,"props":524,"children":525},{},[526],{"type":23,"value":527},"5、魔泊云携手昇思，打造AI解决方案加速企业数智化",{"type":17,"tag":25,"props":529,"children":530},{},[531],{"type":23,"value":532},"在当前的数智化浪潮中，生成式人工智能 (GenAI) 正以前所未有的速度重塑各个行业。作为AI领域的创新先锋，魔泊云（MoPaaS）基于昇思MindSpore AI框架，以DeepSeek和Qwen等国产前沿大模型为基础，为企业提供了切实可行的AI解决方案，帮助企业在人工智能时代实现创新发展。",{"type":17,"tag":25,"props":534,"children":535},{},[536],{"type":17,"tag":116,"props":537,"children":539},{"alt":7,"src":538},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/6dabbdea42b6494ebe63ff895e697f5b.png",[],{"type":17,"tag":25,"props":541,"children":542},{},[543],{"type":17,"tag":29,"props":544,"children":545},{},[546],{"type":23,"value":547},"6、奥看科技携手江苏鲲鹏·昇腾生态创新中心驱动行业全场景跃迁",{"type":17,"tag":25,"props":549,"children":550},{},[551],{"type":23,"value":552},"南京奥看信息科技有限公司携手江苏鲲鹏·昇腾生态创新中心，以昇思MindSpore AI框架为引擎，通过视觉AI大模型与DeepSeek语言模型结合，通过多模态赋能企业生产人、机、料、法、环的全量场景，实现场景的自主监测、智能分析、联动处置和精准治理，解决传统安检及时性差、人力成本高、管理效率低的痛点，助力工业制造企业降本增效、优化生产流程。",{"type":17,"tag":25,"props":554,"children":555},{},[556],{"type":17,"tag":116,"props":557,"children":559},{"alt":7,"src":558},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/bf71c7616b4b42568990e598a204667d.png",[],{"type":17,"tag":25,"props":561,"children":562},{},[563],{"type":17,"tag":29,"props":564,"children":565},{},[566],{"type":23,"value":567},"7、携手昇思，玲珑大模型一体机部署DeepSeek实践",{"type":17,"tag":25,"props":569,"children":570},{},[571],{"type":23,"value":572},"随着DeepSeek在全球持续热度不减，珑京科技联合多方打造DeepSeek大模型的多样化部署解决方案，充分发挥DeepSeek开源模型的低成本高性能优势，进一步推动AI的企业级商用进程。玲珑一体机作为一款开箱即用的成长型AI助手，拥有用户友好界面，操作简单易上手，可节约调优适配时间。其软件包含三大核心产品：智能问答、智能办公和智能报告，支持多模态、多功能、多角度及云端多人并发访问，满足多样化部署需求。",{"type":17,"tag":25,"props":574,"children":575},{},[576],{"type":17,"tag":116,"props":577,"children":579},{"alt":7,"src":578},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/9a4e5e206f5b4d5fbcd97602541f065f.png",[],{"type":17,"tag":25,"props":581,"children":582},{},[583],{"type":17,"tag":29,"props":584,"children":585},{},[586],{"type":23,"value":587},"8、倍特威视基于昇思，以DeepSeek为基础大模型打造异常检测及预测解决方案",{"type":17,"tag":25,"props":589,"children":590},{},[591],{"type":23,"value":592},"倍特威视依托于武汉人工智能计算中心云端算力，离不开昇腾AI硬件强大的算力以及昇思MindSpore AI框架的支撑和稳定运行。通过结合MindIE推理引擎，倍特威视安全生产大模型系统在昇腾硬件上实现了高效且稳定地运行。这一结合使得大模型能够充分利用昇腾的强大算力，为工业安全生产作业中的复杂计算任务提供有力支持。尤其是在面对工业场景隐患排查任务时，即使模型输入的prompt长度达到上万，昇腾配合MindIE推理引擎仍能保持出色的性能表现，确保了模型的稳定性和高效性。",{"type":17,"tag":25,"props":594,"children":595},{},[596],{"type":17,"tag":116,"props":597,"children":599},{"alt":7,"src":598},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/f22d9d1b9df9478cb90c0155683b91fe.png",[],{"type":17,"tag":25,"props":601,"children":602},{},[603],{"type":17,"tag":29,"props":604,"children":605},{},[606],{"type":23,"value":607},"9、星舟志屹基于昇思MindSpore AI框架打造星舟大模型DeepSeek版，加速能源与旅游业智能化升级",{"type":17,"tag":25,"props":609,"children":610},{},[611],{"type":23,"value":612},"随着大模型技术向垂直领域深度渗透，如何在复杂场景中实现高效、精准的智能化应用成为行业关注焦点。西安星舟志屹智能科技有限公司（以下简称“星舟志屹”），作为一家专注于人工智能技术，深耕智慧能源领域应用项目的创新企业，近日正式发布基于昇思MindSpore AI框架 的“星舟大模型DeepSeek版”。该模型通过知识蒸馏、微调技术、私有专家库及海量数据库理解等核心技术，为能源管理提供了全新的解决方案，助力行业实现技术跃迁。",{"type":17,"tag":25,"props":614,"children":615},{},[616],{"type":17,"tag":116,"props":617,"children":619},{"alt":7,"src":618},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/8cfb3356139045d1a8b71bd08c8de6b9.png",[],{"type":17,"tag":25,"props":621,"children":622},{},[623],{"type":17,"tag":29,"props":624,"children":625},{},[626],{"type":23,"value":627},"10、易库智能携手昇思，结合DeepSeek打造大模型知识赋能平台",{"type":17,"tag":25,"props":629,"children":630},{},[631],{"type":23,"value":632},"在数字经济高速发展的今天，知识数据已成为企业发展的\"战略资源\"，如何通过AI技术实现知识的高效管理与价值释放，是各行业数字化转型的核心命题。易库智能基于昇思MindSpore AI框架，成功打造自主创新知识大模型应用体系，通过\"私有化部署+垂直领域适配+知识全链路管理\"的创新模式，为智能办公、装备运维、人才培养等场景提供智能化知识解决方案，助力企业构建智能知识中枢。",{"type":17,"tag":25,"props":634,"children":635},{},[636],{"type":17,"tag":116,"props":637,"children":639},{"alt":7,"src":638},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/b33aa6d37da340ca8361b7cc87b47284.png",[],{"type":17,"tag":25,"props":641,"children":642},{},[643],{"type":17,"tag":29,"props":644,"children":645},{},[646],{"type":23,"value":647},"11、鹏城实验室发布基于昇思框架的强化学习训练框架GRPO-Training-Suite",{"type":17,"tag":25,"props":649,"children":650},{},[651],{"type":23,"value":652},"鹏城实验室基于昇思MindSpore AI框架，率先突破大模型强化学习技术全栈壁垒。通过自主创新硬件、框架及超大规模集群的深度协同，成功构建从硬件算力、算法优化到集群调度的完整技术链条，实现GRPO强化学习训练在Qwen2.5（7B，32B）上的全流程部署，发布并开源GRPO强化学习训练流程和代码，为强化学习开发者提供了训练接口，支持算法快速开发，提供多种训练优化技术。",{"type":17,"tag":25,"props":654,"children":655},{},[656],{"type":23,"value":657},"开源代码仓：",{"type":17,"tag":25,"props":659,"children":660},{},[661],{"type":17,"tag":175,"props":662,"children":665},{"href":663,"rel":664},"https://openi.pcl.ac.cn/PCL-Reasoner/GRPO-Training-Suite",[179],[666],{"type":23,"value":663},{"type":17,"tag":25,"props":668,"children":669},{},[670],{"type":17,"tag":175,"props":671,"children":674},{"href":672,"rel":673},"https://gitee.com/mindspore/mindrlhf",[179],[675],{"type":23,"value":672},{"type":17,"tag":25,"props":677,"children":678},{},[679],{"type":17,"tag":116,"props":680,"children":682},{"alt":7,"src":681},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/6caa43cd70704fe9809cf11df54edb07.png",[],{"type":17,"tag":25,"props":684,"children":685},{},[686],{"type":17,"tag":29,"props":687,"children":688},{},[689],{"type":23,"value":690},"六、技术分享",{"type":17,"tag":25,"props":692,"children":693},{},[694],{"type":17,"tag":29,"props":695,"children":696},{},[697],{"type":23,"value":698},"1、DeepSeek模型Janus-Pro香橙派开发板实践指南",{"type":17,"tag":25,"props":700,"children":701},{},[702],{"type":23,"value":703},"2025 年春节， DeepSeek的Janus-Pro模型横空出世， 为了让更多开发者能够快速上手，昇思社区基于MindSpore AI框架和MindSpore NLP套件在香橙派AI pro（20T）24G开发板上，部署了DeepSeek-Janus-Pro 模型，欢迎广大开发者部署体验。",{"type":17,"tag":25,"props":705,"children":706},{},[707],{"type":23,"value":708},"开源链接：",{"type":17,"tag":25,"props":710,"children":711},{},[712],{"type":17,"tag":175,"props":713,"children":716},{"href":714,"rel":715},"https://gitee.com/mindspore-lab/mindnlp.git",[179],[717],{"type":23,"value":714},{"type":17,"tag":25,"props":719,"children":720},{},[721],{"type":17,"tag":175,"props":722,"children":725},{"href":723,"rel":724},"https://github.com/mindspore-lab/mindnlp.git",[179],[726],{"type":23,"value":723},{"type":17,"tag":25,"props":728,"children":729},{},[730],{"type":17,"tag":116,"props":731,"children":733},{"alt":7,"src":732},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/e884c67caa724c15964775824f93458d.png",[],{"type":17,"tag":25,"props":735,"children":736},{},[737],{"type":17,"tag":29,"props":738,"children":739},{},[740],{"type":23,"value":741},"2、加速DeepSeek训练性能调优，揭秘昇思MindSpore大模型自动流水负载均衡技术",{"type":17,"tag":25,"props":743,"children":744},{},[745],{"type":23,"value":746},"近年来，大模型不断突破参数规模瓶颈，推动模型体量向万亿级跃进，流水线并行(Pipeline Parallelism)已演变为大模型分布式训练的必备技术之一。但当前应用广泛的几种流水线编排技术，如GPipe、PipeDream和Megatron-LM等，在拉高了训练效率的上限的同时也引入了不同设备上内存开销的不均衡，增大了策略调优的难度和成本。",{"type":17,"tag":25,"props":748,"children":749},{},[750],{"type":23,"value":751},"昇思MindSpore团队提出了一种同时均衡计算和内存的全局求解方法，面向复杂的流水线编排逻辑，自动生成最优的stage-layer分配和重计算策略。通过结合昇思MindSpore提供的自动流水负载均衡工具与Dryrun集群内存仿真工具，在大集群上实测DeepSeek-V3 671B调优，实现训练性能倍增；实测千亿参数Llama模型自动负载均衡调优相比于专家调优提升性能14%。",{"type":17,"tag":25,"props":753,"children":754},{},[755],{"type":17,"tag":116,"props":756,"children":758},{"alt":7,"src":757},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/963b861bac7746b3bf7788e29354a5a3.png",[],{"type":17,"tag":25,"props":760,"children":761},{},[762],{"type":17,"tag":29,"props":763,"children":764},{},[765],{"type":23,"value":766},"3、全流程加速MoE模型预训练及强化学习！揭秘昇思MindSpore并行技术",{"type":17,"tag":25,"props":768,"children":769},{},[770],{"type":23,"value":771},"昇思MindSpore已实现DeepSeek V3 + GRPO强化学习训练全流程支持，并将相关代码开源，为强化学习开发者提供了训练接口，支持算法快速开发，提供多种训练优化技术。当前代码已在昇思社区开源，根据本教程，用户可以快速上手体验，或尝试完成DeepSeek V3 R1的训练全过程。",{"type":17,"tag":25,"props":773,"children":774},{},[775,777],{"type":23,"value":776},"开源代码仓地址：",{"type":17,"tag":175,"props":778,"children":781},{"href":779,"rel":780},"https://gitee.com/mindspore/mindrlhf/tree/master/examples/grpo/deepseekv3%5C_grpo%5C_tutorial",[179],[782],{"type":23,"value":783},"https://gitee.com/mindspore/mindrlhf/tree/master/examples/grpo/deepseekv3\\_grpo\\_tutorial",{"type":17,"tag":25,"props":785,"children":786},{},[787],{"type":17,"tag":116,"props":788,"children":790},{"alt":7,"src":789},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/16cbc54782824281bb7a68c548825033.png",[],{"type":17,"tag":25,"props":792,"children":793},{},[794],{"type":17,"tag":29,"props":795,"children":796},{},[797],{"type":23,"value":798},"4、支持GRPO强化学习训练全流程，揭秘MindSpore RLHF套件技术细节",{"type":17,"tag":25,"props":800,"children":801},{},[802],{"type":23,"value":803},"不久之前，MindSpore携手鹏城实验室发布并开源了基于Qwen2.5（7B，32B）模型的GRPO强化学习训练全流程和代码，实现了组件化解耦训练流程与模型定义，通过训推共部署、训练和推理权重在线快速自动重排、异构内存Swap等技术，成功构建从硬件算力、算法优化到集群调度的完整技术链条。本文围绕MindSpore RLHF套件中的关键技术进行分析，重点解读训推共部署、多维混合并行在线快速自动重排、异构内存Swap、支持vLLM等特性。",{"type":17,"tag":25,"props":805,"children":806},{},[807],{"type":17,"tag":116,"props":808,"children":810},{"alt":7,"src":809},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/ff9fc838205b43bb98a7a8876a6370d4.png",[],{"type":17,"tag":25,"props":812,"children":813},{},[814],{"type":17,"tag":29,"props":815,"children":816},{},[817],{"type":23,"value":818},"5、文生视频SOTA模型推理开箱即用！MindSpore支持Step-Video-T2V",{"type":17,"tag":25,"props":820,"children":821},{},[822],{"type":23,"value":823},"开发者基于昇思MindSpore AI框架实现了对阶跃星辰 (stepfun-ai ) 开源的 SOTA文生视频模型Step-Video-T2V支持，并将该版本代码开源，同时完成模型推理支持开箱即用。Step-Video-T2V 具有 30B 参数，能够生成204 帧 544p的高质量视频，在文生视频质量评估 benchmark Step-Video-T2V-Eval上，相比其他开源和或商业模型展示出 SOTA 能力。本文介绍了基于昇思 MindSpore +单机Atlas 800T A2使用Step-Video-T2V的流程。",{"type":17,"tag":25,"props":825,"children":826},{},[827],{"type":23,"value":828},"MindSpore 版 Step-Video-T2V 开源链接：",{"type":17,"tag":25,"props":830,"children":831},{},[832],{"type":17,"tag":175,"props":833,"children":836},{"href":834,"rel":835},"https://github.com/mindspore-lab/mindone/tree/master/examples/step%5C_video%5C_t2v",[179],[837],{"type":23,"value":838},"https://github.com/mindspore-lab/mindone/tree/master/examples/step\\_video\\_t2v",{"type":17,"tag":25,"props":840,"children":841},{},[842],{"type":17,"tag":116,"props":843,"children":845},{"alt":7,"src":844},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/ce1290b3de174c52abfebad60524d8af.png",[],{"type":17,"tag":25,"props":847,"children":848},{},[849],{"type":17,"tag":29,"props":850,"children":851},{},[852],{"type":23,"value":853},"6、视频生成SOTA模型推理开箱即用！MindSpore支持Wan2.1",{"type":17,"tag":25,"props":855,"children":856},{},[857],{"type":23,"value":858},"在AI视频生成领域，Wan2.1作为最新的视觉生成模型，能够根据文本、图像或其他控制信号生成视频，在VBench评测中，Wan2.1以86.22%的总分，摘得桂冠。",{"type":17,"tag":25,"props":860,"children":861},{},[862],{"type":23,"value":863},"MindSpore快速对Wan2.1进行适配，并将其开源至MindSpore ONE仓库，为开发者提供高效体验。本文详细介绍了如何基于昇思MindSpore和单机Atlas 800T A2，完整实现Wan2.1视频生成的部署流程。",{"type":17,"tag":25,"props":865,"children":866},{},[867],{"type":17,"tag":116,"props":868,"children":870},{"alt":7,"src":869},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/a237b477d12b4f38a479e35b7421198b.png",[],{"type":17,"tag":25,"props":872,"children":873},{},[874],{"type":17,"tag":29,"props":875,"children":876},{},[877],{"type":23,"value":878},"7、基于昇思，玩转 CogView4生成汉字！",{"type":17,"tag":25,"props":880,"children":881},{},[882],{"type":23,"value":883},"2025年3月4日，智谱开源了首个支持生成汉字的开源文生图模型 CogView4。该模型具有6B权重，支持原生中文输入，支持中文文字绘画。CogView4 在 DPG-Bench 基准测试中的综合评分排名第一，在开源文生图模型中达到 SOTA 水平。",{"type":17,"tag":25,"props":885,"children":886},{},[887],{"type":23,"value":888},"昇思MindSpore 团队快速支持 CogView4 Diffusers 工作流并开源至MindSpore ONE仓库，结合昇腾硬件，为开发者提供简单易用体验。本文将介绍如何基于昇思MindSpore和单机Atlas 800T A2，通过简单命令，单卡玩转 CogView4 推理。",{"type":17,"tag":25,"props":890,"children":891},{},[892],{"type":23,"value":708},{"type":17,"tag":25,"props":894,"children":895},{},[896],{"type":17,"tag":175,"props":897,"children":900},{"href":898,"rel":899},"https://github.com/mindspore-lab/mindone/tree/master/examples/cogview",[179],[901],{"type":23,"value":898},{"type":17,"tag":25,"props":903,"children":904},{},[905],{"type":23,"value":906},"魔乐社区在线体验：",{"type":17,"tag":25,"props":908,"children":909},{},[910],{"type":17,"tag":175,"props":911,"children":914},{"href":912,"rel":913},"https://modelers.cn/spaces/MindSpore-Lab/cogview4",[179],[915],{"type":23,"value":912},{"type":17,"tag":25,"props":917,"children":918},{},[919],{"type":17,"tag":116,"props":920,"children":922},{"alt":7,"src":921},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/a5f860c7e84a4f468b15f986371ecbd8.png",[],{"type":17,"tag":25,"props":924,"children":925},{},[926],{"type":17,"tag":29,"props":927,"children":928},{},[929],{"type":23,"value":930},"8、昇思MindSpore支持QwQ-32B并上线开源社区",{"type":17,"tag":25,"props":932,"children":933},{},[934],{"type":23,"value":935},"QwQ-32B是阿里云于2025年3月6日发布的人工智能大型语言模型。这是一款拥有 320 亿参数的模型，其性能可与具备 6710 亿参数（其中 370 亿被激活）的 DeepSeek-R1 媲美。",{"type":17,"tag":25,"props":937,"children":938},{},[939],{"type":23,"value":940},"昇思MindSpore原生支持Qwen2.5-32B, 在此基础上0Day完成QwQ-32B的支持，并且完成性能测试。昇思MindSpore开源社区、魔乐社区已第一时间上架该模型，欢迎广大开发者下载体验！",{"type":17,"tag":25,"props":942,"children":943},{},[944,946],{"type":23,"value":945},"魔乐社区：",{"type":17,"tag":175,"props":947,"children":950},{"href":948,"rel":949},"https://modelers.cn/models/MindSpore-Lab/QwQ-32B",[179],[951],{"type":23,"value":948},{"type":17,"tag":25,"props":953,"children":954},{},[955],{"type":17,"tag":116,"props":956,"children":958},{"alt":7,"src":957},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/843eb5d16c094c038c2bf55d29018592.png",[],{"type":17,"tag":25,"props":960,"children":961},{},[962],{"type":17,"tag":29,"props":963,"children":964},{},[965],{"type":23,"value":966},"9、基于MindSpore，玩转HunyuanVideo-I2V图生视频！",{"type":17,"tag":25,"props":968,"children":969},{},[970],{"type":23,"value":971},"HunyuanVideo-I2V是腾讯混元系列模型中一个新型的图像到视频转换模型。用户只需要提供一张图片，简要描述要生成的视频内容，例如动作描述，镜头位置等等，HunyuanVideo-I2V就可以让图片动起来，生成高清连贯的视频，最高支持720P 5s。",{"type":17,"tag":25,"props":973,"children":974},{},[975],{"type":23,"value":976},"昇思MindSpore团队现已完成对HunyuanVideo-I2V的适配，并将其开源至MindSpore ONE仓库，本文详细介绍了如何基于昇思MindSpore和Atlas 800T A2，完整实现HunyuanVideo-I2V从图像到视频生成的部署流程。",{"type":17,"tag":25,"props":978,"children":979},{},[980,982],{"type":23,"value":981},"MindSpore ONE开源代码仓链接：",{"type":17,"tag":175,"props":983,"children":986},{"href":984,"rel":985},"https://github.com/mindspore-lab/mindone/tree/master/examples/hunyuanvideo-i2v",[179],[987],{"type":23,"value":984},{"type":17,"tag":25,"props":989,"children":990},{},[991,993],{"type":23,"value":992},"魔乐社区代码仓链接：",{"type":17,"tag":175,"props":994,"children":997},{"href":995,"rel":996},"https://modelers.cn/models/MindSpore-Lab/hyvideo",[179],[998],{"type":23,"value":995},{"type":17,"tag":25,"props":1000,"children":1001},{},[1002],{"type":17,"tag":116,"props":1003,"children":1005},{"alt":7,"src":1004},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/d91d24cb2a264940bcdff381a92ec4f4.png",[],{"type":17,"tag":25,"props":1007,"children":1008},{},[1009],{"type":17,"tag":29,"props":1010,"children":1011},{},[1012],{"type":23,"value":1013},"10、昇思MindSpore原生论文 | 基于MobileNetV2的家用睡眠呼吸障碍早筛模型",{"type":17,"tag":25,"props":1015,"children":1016},{},[1017],{"type":23,"value":1018},"睡眠呼吸障碍（Sleep Apnea，简称SA）是一种常见但容易被忽视的健康问题，未及时诊断和干预可能会导致心血管疾病、代谢紊乱以及其他健康问题，因此，早期筛查至关重要，但目前的检测主要依赖于多导睡眠图（PSG）测试，复杂、昂贵且需要在专业医疗机构进行，因此开发一种家用的高效、低成本的筛查方法具有重要意义。",{"type":17,"tag":25,"props":1020,"children":1021},{},[1022],{"type":23,"value":1023},"本研究提出了一种基于MobileNetV2的家用睡眠呼吸障碍筛查模型，通过在MindSpore框架上的实现，该模型在保证高预测精度的同时大幅降低了计算成本，且实验结果验证了该模型在实际应用场景中的潜力。",{"type":17,"tag":25,"props":1025,"children":1026},{},[1027],{"type":17,"tag":116,"props":1028,"children":1030},{"alt":7,"src":1029},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/83825c1dc3584359a0e2c5adccfd0f3c.png",[],{"type":17,"tag":25,"props":1032,"children":1033},{},[1034],{"type":17,"tag":29,"props":1035,"children":1036},{},[1037],{"type":23,"value":1038},"11、MindSpore源码精读之ckpt及safetensors权重保存流程",{"type":17,"tag":25,"props":1040,"children":1041},{},[1042],{"type":23,"value":1043},"通过无数次迭代让模型达到满意的性能后，如何保存网络权重这个环节不可忽视。它承载着模型训练的成果与智慧结晶，从模型复用的角度来看，保存的权重能让我们在不同场景下快速调用已训练好的模型，无需重复训练，大大节省时间和计算资源。在迁移学习中，保存的权重更是成为知识迁移的桥梁，助力新任务快速收敛。本文详细介绍了 ckpt 和 safetensors 这两种权重保存的流程。",{"type":17,"tag":25,"props":1045,"children":1046},{},[1047],{"type":17,"tag":116,"props":1048,"children":1050},{"alt":7,"src":1049},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/0d46317438fb4e78aa0f838de49c65eb.png",[],{"type":17,"tag":25,"props":1052,"children":1053},{},[1054],{"type":17,"tag":29,"props":1055,"children":1056},{},[1057],{"type":23,"value":1058},"12、MindSpore源码精读之权重加载实现全解析：从Checkpoint到神经网络的桥梁",{"type":17,"tag":25,"props":1060,"children":1061},{},[1062],{"type":23,"value":1063},"在模型规模、加载速度和兼容性之间，永远存在权衡。深度理解加载流程，本质是掌握框架与硬件、算法与工程的对话规则。MindSpore的权重加载，本质是一场「参数元信息」的精确舞蹈——从.ckpt文件的魔数校验（0x6D73636B），到choice_func的布尔过滤，再到strict_load=False时的后缀匹配（源码中通过_get_parameter_tuple实现），每个环节都暗藏框架设计者对兼容性的极致追求。",{"type":17,"tag":25,"props":1065,"children":1066},{},[1067],{"type":17,"tag":116,"props":1068,"children":1070},{"alt":7,"src":1069},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/a27e06cacdb547148214a0a117f84f1d.png",[],{"type":17,"tag":25,"props":1072,"children":1073},{},[1074],{"type":17,"tag":29,"props":1075,"children":1076},{},[1077],{"type":23,"value":1078},"13、MindSpore Transformers支持DeepSeek-R1蒸馏全流程",{"type":17,"tag":25,"props":1080,"children":1081},{},[1082],{"type":23,"value":1083},"DeepSeek-R1具有生成长思维链（CoT）的能力，这种能力可以通过蒸馏技术迁移到小模型上。小模型学习了R1模型中的知识后，即使不经过强化学习（RL），也能得到出色的推理效果。",{"type":17,"tag":25,"props":1085,"children":1086},{},[1087],{"type":23,"value":1088},"DeepSeek团队通过使用R1模型生成60万条CoT推理数据，蒸馏了Qwen和Llama这样的开源模型，成功验证了这一迁移能力。",{"type":17,"tag":25,"props":1090,"children":1091},{},[1092],{"type":23,"value":1093},"MindSpore Transformers大模型使能套件依托MindSpore AI框架提供的丰富的多维混合并行能力，亲和开源工具与通用数据格式，原生支持大模型蒸馏的端到端全流程开发，提供了高效、便捷的解决方案。",{"type":17,"tag":25,"props":1095,"children":1096},{},[1097],{"type":17,"tag":116,"props":1098,"children":1100},{"alt":7,"src":1099},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/f23a1f0d51a84adc9f580bfa731f6a1d.png",[],{"type":17,"tag":25,"props":1102,"children":1103},{},[1104],{"type":17,"tag":29,"props":1105,"children":1106},{},[1107],{"type":23,"value":1108},"14、视频生成效果惊艳！昇思率先完成Open-Sora 2.0支持",{"type":17,"tag":25,"props":1110,"children":1111},{},[1112],{"type":23,"value":1113},"Open-Sora 2.0是潞晨科技推出的全新开源SOTA视频生成模型，相比传统高性能视频生成模型大幅降低了训练成本。Open-Sora 2.0基于3D自编码器 、3D全注意力机制和MMDiT架构，结合高效的并行训练方案和高压缩比自编码器，显著提升训练效率和推理速度。",{"type":17,"tag":25,"props":1115,"children":1116},{},[1117],{"type":23,"value":1118},"MindSpore团队现已完成对Open-Sora 2.0的适配，并将其开源至MindSpore ONE仓库，本文详细介绍了如何基于昇思MindSpore和单机Atlas 800T A2，完整实现Open-Sora 2.0视频生成流程。",{"type":17,"tag":25,"props":1120,"children":1121},{},[1122,1123],{"type":23,"value":981},{"type":17,"tag":175,"props":1124,"children":1127},{"href":1125,"rel":1126},"https://github.com/mindspore-lab/mindone/tree/master/examples/opensora%5C_hpcai",[179],[1128],{"type":23,"value":1129},"https://github.com/mindspore-lab/mindone/tree/master/examples/opensora\\_hpcai",{"type":17,"tag":25,"props":1131,"children":1132},{},[1133],{"type":17,"tag":116,"props":1134,"children":1136},{"alt":7,"src":1135},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/04/11/e10df2e25a4c4ec3a5b4d642ccba3624.png",[],{"type":17,"tag":25,"props":1138,"children":1139},{},[1140],{"type":17,"tag":29,"props":1141,"children":1142},{},[1143,1145],{"type":23,"value":1144},"七、",{"type":17,"tag":29,"props":1146,"children":1147},{},[1148],{"type":23,"value":1149},"感谢每一位朋友、开发者的支持",{"type":17,"tag":25,"props":1151,"children":1152},{},[1153],{"type":23,"value":1154},"在此感谢社区伙伴们、可爱的小孢子们以及昇思MindSpore SIG组成员们，因为大家的共同努力及辛勤奉献，昇思MindSpore才能不断成长与发展！同时我们对可能出现的不完善之处向您表示诚挚的歉意，并衷心感谢您的理解与支持。",{"type":17,"tag":25,"props":1156,"children":1157},{},[1158],{"type":23,"value":1159},"未来，昇思MindSpore AI框架将持续致力于打造人工智能创新之源，凝聚产业力量，扎根AI根技术，使能大模型与科学智能，成为AI创新的首选框架。",{"title":7,"searchDepth":1161,"depth":1161,"links":1162},4,[],"markdown","content:news:zh:3678.md","content","news/zh/3678.md","news/zh/3678","md",1776506087444]