[{"data":1,"prerenderedAt":245},["ShallowReactive",2],{"content-query-PWcRvoeudd":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"body":13,"_type":239,"_id":240,"_source":241,"_file":242,"_stem":243,"_extension":244},"/news/zh/3765","zh",false,"","加速科研创新，GAITC2025展示多项昇思MindSpore学术基金优秀成果","6月6日-8日，2025全球人工智能技术大会(GAITC2025)在杭州未来科技城举办。","2025-06-18","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/20/b1c91304dbc04cd79a5548cc179ff18a.png","news",{"type":14,"children":15,"toc":236},"root",[16,24,30,38,43,52,57,62,67,72,77,82,87,94,99,107,112,117,124,129,134,141,146,151,158,163,168,175,180,185,192,197,202,209,214,219,226,231],{"type":17,"tag":18,"props":19,"children":21},"element","h1",{"id":20},"加速科研创新gaitc2025展示多项昇思mindspore学术基金优秀成果",[22],{"type":23,"value":8},"text",{"type":17,"tag":25,"props":26,"children":27},"p",{},[28],{"type":23,"value":29},"6月6日-8日，2025全球人工智能技术大会(GAITC2025)在杭州未来科技城举办。本次大会以“交叉、融合、相生、共赢”为主题，汇聚近300位中外专家分享前沿报告，吸引近千名政产学研各界人士参会，共同探讨人工智能技术突破新动态和产业智变新趋势。华为携昇腾AI基础软硬件及多项学术基金优秀项目在大会开幕式、分论坛上重磅亮相，全方位展示了业界基于昇腾平台打造的技术创新以及科研创新、行业应用创新等生态成果。",{"type":17,"tag":25,"props":31,"children":32},{},[33],{"type":17,"tag":34,"props":35,"children":37},"img",{"alt":7,"src":36},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/20/ea5e0c46eff842fc92f49d9d90660832.png",[],{"type":17,"tag":25,"props":39,"children":40},{},[41],{"type":23,"value":42},"2025全球人工智能技术大会(GAITC2025)",{"type":17,"tag":25,"props":44,"children":45},{},[46],{"type":17,"tag":47,"props":48,"children":49},"strong",{},[50],{"type":23,"value":51},"荣誉时刻 学术基金优秀项目颁奖",{"type":17,"tag":25,"props":53,"children":54},{},[55],{"type":23,"value":56},"大会开幕式举行了CAAI-昇思MindSpore/昇腾CANN学术基金优秀项目表彰仪式。中国工程院院士、中国人工智能学会理事长戴琼海和国际欧亚科学院院士田奇为以下六个团队项目颁奖：",{"type":17,"tag":25,"props":58,"children":59},{},[60],{"type":23,"value":61},"中山大学软件工程学院教授陈武辉团队主导的《LongUMoE：基于统一内存的MoE长序列推理框架》",{"type":17,"tag":25,"props":63,"children":64},{},[65],{"type":23,"value":66},"中山大学计算机学院教授、国家优秀青年基金获得者李冠彬团队主导的《基于参数化隐式表达的数字说话人视频生成》",{"type":17,"tag":25,"props":68,"children":69},{},[70],{"type":23,"value":71},"北京安德医智科技有限公司副研究员刘盼团队主导的《基于复合策略的多模态VIT模型推理加速方案的探索》",{"type":17,"tag":25,"props":73,"children":74},{},[75],{"type":23,"value":76},"北京航空航天大学计算机学院助理教授孙庆赟团队主导的《基于MindSpore的PDE基础模型构建研究》",{"type":17,"tag":25,"props":78,"children":79},{},[80],{"type":23,"value":81},"北京交通大学计算机科学与技术学院副教授赵宏智团队主导的《基于CANN平台的高性能Conv算法研究》",{"type":17,"tag":25,"props":83,"children":84},{},[85],{"type":23,"value":86},"天津大学智能与计算学部教授曹兵团队主导《基于CANN平台的无人机视觉大模型》",{"type":17,"tag":25,"props":88,"children":89},{},[90],{"type":17,"tag":34,"props":91,"children":93},{"alt":7,"src":92},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/20/27299b10db944128970930a794e01e84.png",[],{"type":17,"tag":25,"props":95,"children":96},{},[97],{"type":23,"value":98},"CAAI-昇思MindSpore/昇腾CANN 学术基金优秀项目表彰仪式",{"type":17,"tag":25,"props":100,"children":101},{},[102],{"type":17,"tag":47,"props":103,"children":104},{},[105],{"type":23,"value":106},"创新加速 全面展现昇腾AI生态成果",{"type":17,"tag":25,"props":108,"children":109},{},[110],{"type":23,"value":111},"CAAI-昇思MindSpore/昇腾CANN学术基金由CAAI携手华为发起，致力于面向海内外高校及科研院所的AI学者搭建学术交流的平台，提供经费、算力、技术等支持，推动人工智能方法在科学领域的创新应用，持续构建自主创新、开放的科研与人才生态。",{"type":17,"tag":25,"props":113,"children":114},{},[115],{"type":23,"value":116},"以GPT-4为代表的基础模型，彻底变革了语言处理、图像理解与跨模态生成等领域，同时也在向人类智慧的最高结晶“科学研究”发起冲击。例如千禧年问题NS方程如何进行人工智能科学计算求解，是否存在类似于基础模型的Scaling law？北京航空航天大学计算机学院助理教授孙庆赟团队提出了基于MindSpore的面向科学智算的基础模型OmniArch，首个旨在解决多尺度和多物理科学计算问题的基础模型原型，通过全新设计的统一架构解决了问题尺度、物理量、物理规律的三方面难对齐的挑战。该模型在国际基准PDEBench上完成了1D-2D-3D的联合预训练，展现出了较强的上下文学习能力，特别是零样本推理带来了全新物理发现的可能。",{"type":17,"tag":25,"props":118,"children":119},{},[120],{"type":17,"tag":34,"props":121,"children":123},{"alt":7,"src":122},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/20/b362d39857d74ffaa7bab3736897a69a.png",[],{"type":17,"tag":25,"props":125,"children":126},{},[127],{"type":23,"value":128},"北京航空航天大学计算机学院助理教授 孙庆赟团队成果展示",{"type":17,"tag":25,"props":130,"children":131},{},[132],{"type":23,"value":133},"Conv 2d是AI模型中一种常用的算子，其性能直接影响到很多AI模型的整体性能。北京交通大学计算机科学与技术学院副教授赵宏智团队《基于CANN平台的高性能Conv算法研究》项目旨在研究昇腾AI处理器亲和的卷积算子性能优化方法，发现输入特征图的Width维度是优化昇腾处理器片上数据传输效率的一个关键因素，以此提出了一种Width-wise的输入特征图tiling策略，其在不同的输入特征图shape下均能计算出最佳的tiling设置参数，最优化输入特征图tile的片上传输效率，大幅度提升昇腾处理器上卷积算子运行的性能。",{"type":17,"tag":25,"props":135,"children":136},{},[137],{"type":17,"tag":34,"props":138,"children":140},{"alt":7,"src":139},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/20/42353d498bc2415ca9a615c75b1b10f9.png",[],{"type":17,"tag":25,"props":142,"children":143},{},[144],{"type":23,"value":145},"北京交通大学计算机科学与技术学院副教授 赵宏智团队成果展示",{"type":17,"tag":25,"props":147,"children":148},{},[149],{"type":23,"value":150},"在智能无人机低空感知领域，天津大学智能与计算学部教授曹兵团队《基于CANN平台的无人机视觉大模型》项目采集超2000万对低空无人机红外-可见光多模态图像，并基于此构建无人机低空感知基座模型，建立了面向低空无人机视觉的动态可信融合理论体系，提出了基于无人机低空视觉基座模型的多传感器融合算法、低空目标检测-跟踪-计数算法和反无人机算法等全链条技术方案。",{"type":17,"tag":25,"props":152,"children":153},{},[154],{"type":17,"tag":34,"props":155,"children":157},{"alt":7,"src":156},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/20/9b2bcefc01814e628b0b5a29941e1ff5.png",[],{"type":17,"tag":25,"props":159,"children":160},{},[161],{"type":23,"value":162},"天津大学智能与计算学部教授曹兵团队成果",{"type":17,"tag":25,"props":164,"children":165},{},[166],{"type":23,"value":167},"在视觉生成的可控性、个性化和多样性研究上，中山大学计算机学院教授、国家优秀青年基金获得者李冠彬团队开展的《基于参数化隐式表达的数字说话人视频生成》项目，探索了基于MindSpore的多模式驱动视觉内容生成与编辑。团队提出了HiCoDe说话人视频生成算法框架，解决了当前基于扩散模型和神经辐射场(NeRF)的数字人视频生成方法在计算效率和渲染质量上的关键瓶颈问题。HiCoDe采用模块化架构，包括音频编码、可变形3D高斯、渲染合成，在算法效率和性能上较已有方法有大幅提升。",{"type":17,"tag":25,"props":169,"children":170},{},[171],{"type":17,"tag":34,"props":172,"children":174},{"alt":7,"src":173},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/20/2c00f41f73a0445c92057d1a1691f201.png",[],{"type":17,"tag":25,"props":176,"children":177},{},[178],{"type":23,"value":179},"中山大学计算机学院教授、国家优秀青年基金获得者 李冠彬团队成果展示",{"type":17,"tag":25,"props":181,"children":182},{},[183],{"type":23,"value":184},"针对MoE模型由于专家参数规模庞大在推理阶段面临严峻的显存瓶颈问题，中山大学软件工程学院教授陈武辉团队基于MindSpore框架的异构计算能力，进行《LongUMoE：基于统一内存的MoE长序列推理框架》研究，一方面基于专家感知的多batch推理流水线范式，充分利用不同专家间计算与I/O特性差异，有效压缩层间与层内气泡；另一方面利用相邻门控层输入间的高余弦相似度进行跨层预取，从而实现高预测精度而无需额外的GPU开销，最终实现MoE模型在资源受限场景下的高吞吐推理。",{"type":17,"tag":25,"props":186,"children":187},{},[188],{"type":17,"tag":34,"props":189,"children":191},{"alt":7,"src":190},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/20/043cb184048b49d486dbc1f396a81dff.png",[],{"type":17,"tag":25,"props":193,"children":194},{},[195],{"type":23,"value":196},"中山大学软件工程学院教授陈武辉团队成果展示",{"type":17,"tag":25,"props":198,"children":199},{},[200],{"type":23,"value":201},"为解决当前模型在高清三维医学影像场景显存占用高、推理耗时长等实际应用难题，北京安德医智科技有限公司副研究员刘盼团队依托MindSpore框架，探索基于复合策略的多模态VIT模型推理加速方案。通过额外的轻量级路由网络，将原始图像的下采样特征作为输入，输出每个Patch的连续重要性分数。训练阶段，在保持原有SAM分割精度的基础上，通过联合“分割损失＋稀疏正则”让路由网络学会剔除冗余区域。推理时，仅对得分高于阈值的Patch进行完整ViT编码，其余区域直接由SAM解码器通过位置插值等方式补全，从而有效提升模型在资源受限环境下的推理效率，降低部署门槛，为AI模型在基层医疗场景中的落地应用提供高性能、低成本的解决方案。",{"type":17,"tag":25,"props":203,"children":204},{},[205],{"type":17,"tag":34,"props":206,"children":208},{"alt":7,"src":207},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/20/0afea4425a504873a7c3f89d74ed733f.png",[],{"type":17,"tag":25,"props":210,"children":211},{},[212],{"type":23,"value":213},"北京安德医智科技有限公司副研究员刘盼团队成果展示",{"type":17,"tag":25,"props":215,"children":216},{},[217],{"type":23,"value":218},"此外，大会还展示了《基于CANN平台的高性能融合Kernel算法研究与优化》、《基于多模态的第一视角视频行为分析大模型研究》、《基于多模态的人体感知大模型研究》、《基于脉冲神经网络的大语言模型架构研究》、《基于人工智能方法的东亚高分辨率气候数据重建》、《基于多组学计算的药物重定位方法》、《基于三维场景高效表征的多模态大语言模型研究》、《基于MindSpore的大模型训练高效参数微调、Prompt-tuning指令学习与强化学习(RLHF)》、《基于MindFormer和MindCV的视觉一语言模型的上下文学习框架》学术基金项目科研成果。",{"type":17,"tag":25,"props":220,"children":221},{},[222],{"type":17,"tag":34,"props":223,"children":225},{"alt":7,"src":224},"https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/20/91441ef5df754c01ae1664b51809fa4a.png",[],{"type":17,"tag":25,"props":227,"children":228},{},[229],{"type":23,"value":230},"2024年学术基金项目成果展示",{"type":17,"tag":25,"props":232,"children":233},{},[234],{"type":23,"value":235},"以学术基金为后盾，以论坛交流为契机，昇腾与中国人工智能学会的深度合作，探索出了一条产学研协同攻关、联合创新的可行路径。目前，2025年度中国人工智能学会-昇思MindSpore学术基金、昇腾CANN学术基金项目正在开放申报中，欢迎广大科研工作者申报，具体可登录中国人工智能学会官网查看申报指南与常见问题。",{"title":7,"searchDepth":237,"depth":237,"links":238},4,[],"markdown","content:news:zh:3765.md","content","news/zh/3765.md","news/zh/3765","md",1776506089118]