[{"data":1,"prerenderedAt":683},["ShallowReactive",2],{"content-query-cLR75SdSbO":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"body":13,"_type":677,"_id":678,"_source":679,"_file":680,"_stem":681,"_extension":682},"/news/zh/3721","zh",false,"","社区月报 | 昇思MindSpore 2.6版本重磅发布、Qwen3、GLM4等主流模型0Day同步首发！","昇思MindSpore2025年4月份总结","2025-05-13","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/e7c80b5bec8f4440ba805a27fc1c7e5a.png","news",{"type":14,"children":15,"toc":674},"root",[16,24,37,45,50,55,60,65,70,75,80,88,93,98,117,125,135,140,145,153,161,166,171,176,181,192,197,206,213,223,228,235,245,250,257,267,272,279,286,294,302,307,312,323,330,338,343,350,359,364,369,380,387,396,401,408,417,422,429,437,446,451,456,463,473,478,485,494,499,506,514,523,528,533,538,548,557,562,567,576,583,592,597,602,611,618,627,639,651,664,669],{"type":17,"tag":18,"props":19,"children":21},"element","h1",{"id":20},"社区月报-昇思mindspore-26版本重磅发布qwen3glm4等主流模型0day同步首发",[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4月，昇思MindSpore开源社区持续保持着高速发展的态度，在技术创新、社区运作等方面都取得了显著的成果。",{"type":17,"tag":25,"props":51,"children":52},{},[53],{"type":23,"value":54},"社区同步首发Qwen3，支持一键部署、0Day支持智谱开源GLM-4-0414系列模型并上线MindSpore社区，GLM4涵盖32B和9B规格，推理速度高达200 Tokens/秒，进一步丰富了开源模型生态。上海交通大学联合MindSpore与openEuler社区，成功实现DeepSeek全栈开源单机推理部署，在量化精度无损的条件下显著提升推理性能。",{"type":17,"tag":25,"props":56,"children":57},{},[58],{"type":23,"value":59},"昇思开发者大会（MindSpore Developer Day 2025）成功举办，正式发布MindSpore 2.6版本，该版本全面支持DeepSeek V3/R1 MoE模型训练推理全流程，并推出强化学习套件，显著提升模型迁移效率。",{"type":17,"tag":25,"props":61,"children":62},{},[63],{"type":23,"value":64},"在创新应用方面，中国科学技术大学团队提出了一种名为ChemAgents的多智能体系统，结合了LLM和自动化机器人技术，能够自主执行复杂的化学实验，极大地减少了人为干预。该研究使用的是基于昇思MindSpore的Llama-3.1-70B模型，目前已昇思MindSpore社区开源。",{"type":17,"tag":25,"props":66,"children":67},{},[68],{"type":23,"value":69},"华为阔折叠屏手机Pura X搭载全新小艺助手，全新小艺基于昇思MindSpore，语音合成性能提升30%，AI人像修图效率倍增，带来更拟人化的自然语音交互体验。翔迅科技则基于昇腾和MindSpore推出DeepSeek智能办公大模型，推动行业智能化转型。",{"type":17,"tag":25,"props":71,"children":72},{},[73],{"type":23,"value":74},"在社区活动方面，昇思MindSpore社区通过创新应用挑战、华为ICT大赛、开源实习计划、量子计算黑客松大赛等活动推动AI技术在不同场景中的深度应用，助力开发者成长。值得一提的是，在第九届华为ICT大赛中，基于MindSpore开发的创新项目表现亮眼，斩获半数一等奖，展现了社区开发者的技术实力。在开发者培养方面，昇思MindSpore开源实习项目启动，在校学生可通过完成实际任务获得积分和实习证明。高校合作方面，昇思大模型创新应用与工程实践师资培训成功举办，吸引了全国32所高校教师参与，推动AI技术在教学中的创新应用。",{"type":17,"tag":25,"props":76,"children":77},{},[78],{"type":23,"value":79},"以下是4月昇思MindSpore社区进展的详细报告。",{"type":17,"tag":25,"props":81,"children":82},{},[83],{"type":17,"tag":29,"props":84,"children":85},{},[86],{"type":23,"value":87},"二、社区规模",{"type":17,"tag":25,"props":89,"children":90},{},[91],{"type":23,"value":92},"截至4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":94,"children":95},{},[96],{"type":23,"value":97},"社区发展了**4.6万+**核心贡献者；",{"type":17,"tag":25,"props":99,"children":100},{},[101,103,108,110,115],{"type":23,"value":102},"累计产生ISSUE共",{"type":17,"tag":29,"props":104,"children":105},{},[106],{"type":23,"value":107},"41.6****K",{"type":23,"value":109},"、PR共",{"type":17,"tag":29,"props":111,"children":112},{},[113],{"type":23,"value":114},"110.5K",{"type":23,"value":116},"。",{"type":17,"tag":25,"props":118,"children":119},{},[120],{"type":17,"tag":29,"props":121,"children":122},{},[123],{"type":23,"value":124},"三、社区大事件",{"type":17,"tag":25,"props":126,"children":127},{},[128,130],{"type":23,"value":129},"**1、**",{"type":17,"tag":29,"props":131,"children":132},{},[133],{"type":23,"value":134},"昇思开发者大会成功举办，MindSpore 2.6版本重磅发布",{"type":17,"tag":25,"props":136,"children":137},{},[138],{"type":23,"value":139},"2025年4月12日，昇思开发者大会（MindSpore Developer Day 2025）顺利召开，吸引了超过300位开发者现场深度参与。会上，昇思MindSpore 2.6版本正式发布。该版本全面支持类DeepSeek V3/R1 MoE模型训练推理全流程，推出训推一体的强化学习套件使能后训练范式创新，易用性上实现主流SOTA模型的Day0迁移。",{"type":17,"tag":25,"props":141,"children":142},{},[143],{"type":23,"value":144},"在SIG Gathering环节，开发者们围绕《MindSpore MOE专家大模型预训练技术研讨》、《MindSpore强化学习、蒸馏与R1复现技术研讨》、《MindSpore加速DeepSeek大模型推理研讨》、《MindSpore开发体验与易用性提升》、《MindSpore Science：加速AI4Science创新与落地》五大专题展开深入探讨与交流，推动昇思MindSpore的技术发展与应用。",{"type":17,"tag":25,"props":146,"children":147},{},[148],{"type":17,"tag":149,"props":150,"children":152},"img",{"alt":7,"src":151},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/aed8f09496f4430ea02b6b7a05b93f5d.png",[],{"type":17,"tag":25,"props":154,"children":155},{},[156],{"type":17,"tag":29,"props":157,"children":158},{},[159],{"type":23,"value":160},"2、同****步首发！智谱开源GLM-4-0414系列6个模型上线昇思MindSpore",{"type":17,"tag":25,"props":162,"children":163},{},[164],{"type":23,"value":165},"4月15日，智谱全新开源的GLM-4-0414 32B/9B模型上线昇思MindSpore开源社区，并完成精度测试。",{"type":17,"tag":25,"props":167,"children":168},{},[169],{"type":23,"value":170},"智谱GLM-4-0414拥有 32B/9B两个尺寸，涵盖基座、推理、沉思模型，均遵循 MIT 许可协议。其中，推理模型 GLM-Z1-32B-0414 性能媲美 DeepSeek-R1 等顶尖模型，实测推理速度可达 200 Tokens/秒。",{"type":17,"tag":25,"props":172,"children":173},{},[174],{"type":23,"value":175},"基座模型 GLM-4-32B-0414 拥有 320 亿参数，其性能可与国内、外参数量更大的主流模型相媲美。该模型利用 15T 高质量数据进行预训练，特别纳入了丰富的推理类合成数据，为后续的强化学习扩展奠定了基础。在后训练阶段，除了进行面向对话场景的人类偏好对齐，还通过拒绝采样和强化学习等技术，重点增强模型在指令遵循、工程代码生成、函数调用等任务上的表现，以强化智能体任务所需的原子能力。",{"type":17,"tag":25,"props":177,"children":178},{},[179],{"type":23,"value":180},"昇思MindSpore开源社区训练代码：",{"type":17,"tag":25,"props":182,"children":183},{},[184],{"type":17,"tag":185,"props":186,"children":190},"a",{"href":187,"rel":188},"https://gitee.com/mindspore/mindformers/tree/dev/research/deepseek3",[189],"nofollow",[191],{"type":23,"value":187},{"type":17,"tag":25,"props":193,"children":194},{},[195],{"type":23,"value":196},"魔乐社区推理代码：",{"type":17,"tag":25,"props":198,"children":199},{},[200],{"type":17,"tag":185,"props":201,"children":204},{"href":202,"rel":203},"https://modelers.cn/models/MindSpore-Lab/DeepSeek-V3",[189],[205],{"type":23,"value":202},{"type":17,"tag":25,"props":207,"children":208},{},[209],{"type":17,"tag":149,"props":210,"children":212},{"alt":7,"src":211},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/5fce97be94694348b5cf22523e66a7e7.png",[],{"type":17,"tag":25,"props":214,"children":215},{},[216,218],{"type":23,"value":217},"**3、**",{"type":17,"tag":29,"props":219,"children":220},{},[221],{"type":23,"value":222},"上海交通大学联合MindSpore与openEuler社区，实现DeepSeek全栈开源单机推理部署",{"type":17,"tag":25,"props":224,"children":225},{},[226],{"type":23,"value":227},"上海交通大学并行与分布式系统研究所联合昇思MindSpore社区、openEuler社区，联合开发验证了基于vLLM+openEuler+MindSpore、全栈开源的DeepSeek推理解决方案。双方研发人员协同完成DeepSeek-R1/V3-0324权重参数的Int4量化，在精度几乎无损的条件下，实现单台Atlas 800T A2服务器（64GB）部署运行DeepSeek大模型推理服务。同时，验证了DeepSeek-R1/V3-0324 Int8量化推理，其系统总吞吐性能较前序版本提升幅度达15%。",{"type":17,"tag":25,"props":229,"children":230},{},[231],{"type":17,"tag":149,"props":232,"children":234},{"alt":7,"src":233},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/fa1c6979f6bf40189df4556592f18cc3.png",[],{"type":17,"tag":25,"props":236,"children":237},{},[238,240],{"type":23,"value":239},"**4、**",{"type":17,"tag":29,"props":241,"children":242},{},[243],{"type":23,"value":244},"0 Day支持！昇思MindSpore同步首发Qwen3，支持一键部署",{"type":17,"tag":25,"props":246,"children":247},{},[248],{"type":23,"value":249},"Qwen3是阿里云于2025年4月29日发布并开源的全新模型，作为Qwen系列中的最新一代大语言模型，提供了一系列密集型和混合专家（MoE）模型。本次Qwen发布多个尺寸模型，覆盖235B/32B/30B/14B/8B/4B/1.7B/0.6B。昇思MindSpore基于对Qwen2.5的支持与兼容主流生态的接口，快速实现Qwen3的0Day支持，并将MindSpore版Qwen3代码上传至开源社区代码仓，面向开发者提供开箱即用的模型。",{"type":17,"tag":25,"props":251,"children":252},{},[253],{"type":17,"tag":149,"props":254,"children":256},{"alt":7,"src":255},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/bf077fa0e5af44c5a6fcf05783958d4e.png",[],{"type":17,"tag":25,"props":258,"children":259},{},[260,262],{"type":23,"value":261},"**5、**",{"type":17,"tag":29,"props":263,"children":264},{},[265],{"type":23,"value":266},"昇思开源社区携手鹏城实验室荣获第四届中国国际软件发展大会突破性技术成果奖",{"type":17,"tag":25,"props":268,"children":269},{},[270],{"type":23,"value":271},"2025年4月24日，由中国软件行业协会主办的第四届中国国际软件发展大会在京召开。会上，昇思MindSpore AI框架全新发布的MindSpore 2.6版本、鹏城实验室发布的基于昇思MindSpore AI框架及昇腾AI硬件的强化学习训练框架GRPO-Training-Suite在本次大会上均荣获2024年软件行业突破性技术成果奖。",{"type":17,"tag":25,"props":273,"children":274},{},[275],{"type":17,"tag":149,"props":276,"children":278},{"alt":7,"src":277},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/d32c143c8ef24ce3bf5b0ac1dfd440c3.png",[],{"type":17,"tag":25,"props":280,"children":281},{},[282],{"type":17,"tag":149,"props":283,"children":285},{"alt":7,"src":284},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/e9dff016467d4d72925a5d7e06d12e7e.png",[],{"type":17,"tag":25,"props":287,"children":288},{},[289],{"type":17,"tag":29,"props":290,"children":291},{},[292],{"type":23,"value":293},"四、社区动态",{"type":17,"tag":25,"props":295,"children":296},{},[297],{"type":17,"tag":29,"props":298,"children":299},{},[300],{"type":23,"value":301},"1、创新应用挑战活动第一期开启报名",{"type":17,"tag":25,"props":303,"children":304},{},[305],{"type":23,"value":306},"模力方舟携手昇腾与昇思MindSpore，重磅推出「昇腾赋能，智启未来」创新应用挑战！挑战活动第一期报名已正式开启，参与的应用将聚焦于学习效能提升与教育场景革新，基于昇腾与昇思MindSpore支持的模型开发具备实用价值与创新性的教育类应用。",{"type":17,"tag":25,"props":308,"children":309},{},[310],{"type":23,"value":311},"欢迎使用昇思MindSpore模型参赛",{"type":17,"tag":25,"props":313,"children":314},{},[315,317],{"type":23,"value":316},"模型链接：",{"type":17,"tag":185,"props":318,"children":321},{"href":319,"rel":320},"https://ai.gitee.com/serverless-api/packages/1495",[189],[322],{"type":23,"value":319},{"type":17,"tag":25,"props":324,"children":325},{},[326],{"type":17,"tag":149,"props":327,"children":329},{"alt":7,"src":328},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/cadc898a6d284052a99c79a1c19e6787.png",[],{"type":17,"tag":25,"props":331,"children":332},{},[333],{"type":17,"tag":29,"props":334,"children":335},{},[336],{"type":23,"value":337},"2、第九届华为ICT大赛创新赛中国总决赛落幕，昇思开发者斩获半数一等奖",{"type":17,"tag":25,"props":339,"children":340},{},[341],{"type":23,"value":342},"第九届华为ICT大赛中国总决赛颁奖典礼在东南大学圆满落下帷幕。本次决赛各赛道获奖名单在会上揭晓。在创新赛，昇思MindSpore开发者脱颖而出，占一等奖获奖队伍的50%，他们将晋级5月在深圳举办的全球总决赛，与来自世界其他国家的队伍展开巅峰对决。",{"type":17,"tag":25,"props":344,"children":345},{},[346],{"type":17,"tag":149,"props":347,"children":349},{"alt":7,"src":348},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/aba4904ee78d44709599635ab47fc169.png",[],{"type":17,"tag":25,"props":351,"children":352},{},[353,354],{"type":23,"value":217},{"type":17,"tag":29,"props":355,"children":356},{},[357],{"type":23,"value":358},"第七届·2025 MindSpore量子计算黑客松全国大赛火热进行中",{"type":17,"tag":25,"props":360,"children":361},{},[362],{"type":23,"value":363},"由昇思MindSpore社区主办的年度量子计算黑客松全国大赛正在火热进行中，诚邀您报名。无论您是初学小白，还是技术达人，参赛即有收获，欢迎参赛。",{"type":17,"tag":25,"props":365,"children":366},{},[367],{"type":23,"value":368},"活动报名截止日期：5月20日",{"type":17,"tag":25,"props":370,"children":371},{},[372,374],{"type":23,"value":373},"大赛报名链接：",{"type":17,"tag":185,"props":375,"children":378},{"href":376,"rel":377},"https://developer.huaweicloud.com/competition/information/1300000041",[189],[379],{"type":23,"value":376},{"type":17,"tag":25,"props":381,"children":382},{},[383],{"type":17,"tag":149,"props":384,"children":386},{"alt":7,"src":385},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/f892d163f1c240839660ed177dccd64e.png",[],{"type":17,"tag":25,"props":388,"children":389},{},[390,391],{"type":23,"value":239},{"type":17,"tag":29,"props":392,"children":393},{},[394],{"type":23,"value":395},"做任务攒积分，昇思MindSpore开源实习活动持续进行中",{"type":17,"tag":25,"props":397,"children":398},{},[399],{"type":23,"value":400},"开源实习是由MindSpore社区、openEuler社区等多个社区共同发起的线上实习项目，旨在鼓励在校学生积极参与开源社区，在实际的开源环境中提升实践能力。昇思MindSpore社区将持续提供实习任务与导师辅导，学生通过实习申请后，可在社区领取任务，每完成一个任务可获得相应积分，积分累计达规定量后，可获得实习证明和实习工资。欢迎感兴趣的开发者参加。",{"type":17,"tag":25,"props":402,"children":403},{},[404],{"type":17,"tag":149,"props":405,"children":407},{"alt":7,"src":406},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/81b8cc3ed8c34a1bb1d1b8210c0305a0.png",[],{"type":17,"tag":25,"props":409,"children":410},{},[411,412],{"type":23,"value":261},{"type":17,"tag":29,"props":413,"children":414},{},[415],{"type":23,"value":416},"2025年华为基础软件师资培训：昇思大模型创新应用与工程实践",{"type":17,"tag":25,"props":418,"children":419},{},[420],{"type":23,"value":421},"2025年4月26-27日， 2025年华为基础软件-昇思大模型创新应用与工程实践师资培训在北京师范大学成功举办，本次培训吸引了32位来自全国多所高校计算机科学与技术、人工智能等相关专业的教师参与，共同探索大模型技术在教学与科研中的创新应用。 ",{"type":17,"tag":25,"props":423,"children":424},{},[425],{"type":17,"tag":149,"props":426,"children":428},{"alt":7,"src":427},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/13dfeca5d81241f6975e36510dfea0a7.png",[],{"type":17,"tag":25,"props":430,"children":431},{},[432],{"type":17,"tag":29,"props":433,"children":434},{},[435],{"type":23,"value":436},"五、社区案例",{"type":17,"tag":25,"props":438,"children":439},{},[440,441],{"type":23,"value":129},{"type":17,"tag":29,"props":442,"children":443},{},[444],{"type":23,"value":445},"昇思助力中科大团队打造多智能体驱动的机器人AI化学家",{"type":17,"tag":25,"props":447,"children":448},{},[449],{"type":23,"value":450},"由中国科学技术大学罗毅、陈林江、张飞、尚伟伟、傅尧和江俊等组成的团队在《J. Am. Chem. Soc.》提出了一种名为ChemAgents的多智能体系统，结合了LLM和自动化机器人技术，能够自主执行复杂的化学实验，极大地减少了人为干预。通过多智能体的协作，ChemAgents不仅能够自主设计实验、执行实验，还能通过机器学习模型进行数据分析和优化，最终实现化学研究的全自动化。",{"type":17,"tag":25,"props":452,"children":453},{},[454],{"type":23,"value":455},"据了解，该研究使用的是基于昇思MindSpore的Llama-3.1-70B模型，目前已昇思MindSpore社区开源。",{"type":17,"tag":25,"props":457,"children":458},{},[459],{"type":17,"tag":149,"props":460,"children":462},{"alt":7,"src":461},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/79bc0011d584477c8db875c754c7d1f8.png",[],{"type":17,"tag":25,"props":464,"children":465},{},[466,468],{"type":23,"value":467},"**2、**",{"type":17,"tag":29,"props":469,"children":470},{},[471],{"type":23,"value":472},"算法性能倍增！昇思多维混合并行等技术助力小艺语音交互、AI修图体验升级",{"type":17,"tag":25,"props":474,"children":475},{},[476],{"type":23,"value":477},"近期，华为发布阔折叠屏手机Pura X，并首发鸿蒙AI，搭载焕新升级的小艺助手，带来更拟人化的自然语音交互体验。全新小艺基于昇思MindSpore，实现小艺语音合成算法推理性能提升30%，AI人像精修算法推理性能倍增，提升用户对话体验的实时性与修图效率。",{"type":17,"tag":25,"props":479,"children":480},{},[481],{"type":17,"tag":149,"props":482,"children":484},{"alt":7,"src":483},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/2cb5fb19395f480f8c758c3b8cadc52c.png",[],{"type":17,"tag":25,"props":486,"children":487},{},[488,489],{"type":23,"value":217},{"type":17,"tag":29,"props":490,"children":491},{},[492],{"type":23,"value":493},"翔迅科技基于昇腾和昇思MindSpore打造翔迅大模型DeepSeek智能办公版，助力行业办公场景AI赋能",{"type":17,"tag":25,"props":495,"children":496},{},[497],{"type":23,"value":498},"随着人工智能技术的极速发展，如何在复杂的办公场景中实现高效、精准的智能化应用，已成为业界关注的焦点。西安翔迅科技有限责任公司（简称“翔迅科技”），以多年的技术经验和强大的人工智能技术团队，在西安昇腾智能科技有限公司的AI算力支持下，基于昇腾AI硬件+昇思MindSpore AI框架，打造出“翔迅大模型DeepSeek智能办公版”。通过知识蒸馏、定制化微调、垂直领域知识库及大规模数据处理等技术，为办公场景提供了新的解决方案，助力行业办公数字化、智能化转型。",{"type":17,"tag":25,"props":500,"children":501},{},[502],{"type":17,"tag":149,"props":503,"children":505},{"alt":7,"src":504},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/d545c408e0214cec90b2e85e879cf93b.png",[],{"type":17,"tag":25,"props":507,"children":508},{},[509],{"type":17,"tag":29,"props":510,"children":511},{},[512],{"type":23,"value":513},"六、技术分享",{"type":17,"tag":25,"props":515,"children":516},{},[517,518],{"type":23,"value":129},{"type":17,"tag":29,"props":519,"children":520},{},[521],{"type":23,"value":522},"快速支持DeepSeek-V3-0324，昇思MindSpore+vLLM服务化部署开源版本上线",{"type":17,"tag":25,"props":524,"children":525},{},[526],{"type":23,"value":527},"近日，基于昇思 MindSpore AI 框架+vLLM的DeepSeek-V3-0324 模型完成适配并上线魔乐社区。",{"type":17,"tag":25,"props":529,"children":530},{},[531],{"type":23,"value":532},"作为开源AI框架，MindSpore 以其卓越的性能优化、灵活的模型开发能力和高效的分布式训练能力，为大规模模型训推部署提供全流程开源支持。",{"type":17,"tag":25,"props":534,"children":535},{},[536],{"type":23,"value":537},"原文阅读：",{"type":17,"tag":25,"props":539,"children":540},{},[541],{"type":17,"tag":185,"props":542,"children":545},{"href":543,"rel":544},"https://mp.weixin.qq.com/s?__biz=MzkxMTM2MjMzNg==&mid=2247631070&idx=2&sn=203e48ddd377b9dd29b7229134bc2ac2&scene=21#wechat_redirect",[189],[546],{"type":23,"value":547},"https://mp.weixin.qq.com/s/DfEc14anPJ6_vO_OPJNOdA",{"type":17,"tag":25,"props":549,"children":550},{},[551,552],{"type":23,"value":467},{"type":17,"tag":29,"props":553,"children":554},{},[555],{"type":23,"value":556},"零样本声音克隆，基于昇腾+MindSpore玩转Spark-TTS",{"type":17,"tag":25,"props":558,"children":559},{},[560],{"type":23,"value":561},"Spark-TTS 是一款基于大语言模型（LLM）技术的先进文本转语音系统，能够根据用户需求合成高准确度且自然流畅的定制化语音。MindSpore团队现已完成对Spark-TTS 的适配，并将其开源至MindSpore ONE仓库。",{"type":17,"tag":25,"props":563,"children":564},{},[565],{"type":23,"value":566},"MindSpore ONE开源代码仓链接：",{"type":17,"tag":25,"props":568,"children":569},{},[570],{"type":17,"tag":185,"props":571,"children":574},{"href":572,"rel":573},"https://github.com/mindspore-lab/mindone/tree/master/examples/sparktts",[189],[575],{"type":23,"value":572},{"type":17,"tag":25,"props":577,"children":578},{},[579],{"type":17,"tag":149,"props":580,"children":582},{"alt":7,"src":581},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/690a4546817043e1a2088cf775a818b9.png",[],{"type":17,"tag":25,"props":584,"children":585},{},[586,587],{"type":23,"value":217},{"type":17,"tag":29,"props":588,"children":589},{},[590],{"type":23,"value":591},"晶体生成10倍加速！基于流模型的晶体结构生成模型CrystalFlow在昇思社区开源",{"type":17,"tag":25,"props":593,"children":594},{},[595],{"type":23,"value":596},"理论晶体结构预测是通过计算的手段寻找物质在给定的外界条件下最稳定结构的重要手段。传统方法依赖势能面的随机采样与局域优化，计算成本高昂，尤其在多元素复杂体系中面临巨大挑战。CrystalFlow采用先进的流模型生成方案，有效地在未知的目标概率分布和已知的简单概率分布建立概率密度变换关系，以学习晶体结构数据库中合理的稳定结构特征，并生成新的合理构型。通过使用图等变消息传递网络，CrystalFlow有效处理了晶体结构内在的基本对称性，包括置换、旋转和周期平移不变性，实现对晶体结构的高效学习和建模，以及灵活的条件式生成策略。该模型在MP20等基准数据集上达到优秀的水平，目前已在昇思社区开源。",{"type":17,"tag":25,"props":598,"children":599},{},[600],{"type":23,"value":601},"项目地址：",{"type":17,"tag":25,"props":603,"children":604},{},[605],{"type":17,"tag":185,"props":606,"children":609},{"href":607,"rel":608},"https://gitee.com/mindspore/mindscience/tree/master/MindChemistry",[189],[610],{"type":23,"value":607},{"type":17,"tag":25,"props":612,"children":613},{},[614],{"type":17,"tag":149,"props":615,"children":617},{"alt":7,"src":616},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/05/16/64791e61a2404fa3bedf7343245ac357.png",[],{"type":17,"tag":25,"props":619,"children":620},{},[621,622],{"type":23,"value":239},{"type":17,"tag":29,"props":623,"children":624},{},[625],{"type":23,"value":626},"社区博客分享（4月）",{"type":17,"tag":25,"props":628,"children":629},{},[630,632],{"type":23,"value":631},"1）",{"type":17,"tag":185,"props":633,"children":636},{"href":634,"rel":635},"https://mp.weixin.qq.com/s?__biz=MzkxMTM2MjMzNg==&mid=2247631376&idx=1&sn=cb1fe79d4c8906f4beba0201eafffb37&scene=21#wechat_redirect",[189],[637],{"type":23,"value":638},"昇思MindSpore混合精度训练源码大揭秘",{"type":17,"tag":25,"props":640,"children":641},{},[642,644],{"type":23,"value":643},"2）",{"type":17,"tag":185,"props":645,"children":648},{"href":646,"rel":647},"https://mp.weixin.qq.com/s?__biz=MzkxMTM2MjMzNg==&mid=2247631401&idx=1&sn=e7a0673e4ae08508a9c5345e3b00884e&scene=21#wechat_redirect",[189],[649],{"type":23,"value":650},"源码精读之MindSpore如何实现Python侧和C++侧之间的链接",{"type":17,"tag":25,"props":652,"children":653},{},[654],{"type":17,"tag":29,"props":655,"children":656},{},[657,659],{"type":23,"value":658},"七、",{"type":17,"tag":29,"props":660,"children":661},{},[662],{"type":23,"value":663},"感谢每一位朋友、开发者的支持",{"type":17,"tag":25,"props":665,"children":666},{},[667],{"type":23,"value":668},"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":670,"children":671},{},[672],{"type":23,"value":673},"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":675,"depth":675,"links":676},4,[],"markdown","content:news:zh:3721.md","content","news/zh/3721.md","news/zh/3721","md",1776506088374]