[{"data":1,"prerenderedAt":415},["ShallowReactive",2],{"content-query-PV4jy48iYb":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"body":13,"_type":409,"_id":410,"_source":411,"_file":412,"_stem":413,"_extension":414},"/news/zh/3378","zh",false,"","AI创新之旅，昇思MindSpore创新训练营·长三角站成功举办","8月27日-30日，昇思MindSpore创新训练营·长三角站在上海交通大学顺利举行。","2024-09-05","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2024/09/06/165cc15c275745118d22d382ae3e9072.png","news",{"type":14,"children":15,"toc":406},"root",[16,24,30,39,47,52,57,65,72,77,82,101,106,111,118,131,139,144,149,156,161,166,173,177,186,194,199,204,211,226,231,236,241,246,251,258,262,269,276,284,289,293,300,310,315,320,325,330,335,340,345,350,355,362,367,372,379,384,389,396,401],{"type":17,"tag":18,"props":19,"children":21},"element","h1",{"id":20},"ai创新之旅昇思mindspore创新训练营长三角站成功举办",[22],{"type":23,"value":8},"text",{"type":17,"tag":25,"props":26,"children":27},"p",{},[28],{"type":23,"value":29},"8月27日-30日，昇思MindSpore创新训练营·长三角站在上海交通大学顺利举行。本次活动由上海交通大学、华为技术有限公司、昇思MindSpore开源社区主办，长三角各高校教务处协办，吸引了来自上海交通大学、浙江大学、复旦大学、南京大学、中国科学技术大学、同济大学、华东师范大学、华东理工大学、中国石油大学（北京）、南京邮电大学、南京林业大学、加州大学洛杉矶分校共12所高校的79名学生参与。经过连续几天的学习与实践，学员们不仅对人工智能及大模型领域的最新进展进行了深入探讨，还全面了解了行业发展的最新趋势，为自己在未来科技创新的道路上奠定了坚实的基础。",{"type":17,"tag":25,"props":31,"children":32},{},[33],{"type":17,"tag":34,"props":35,"children":38},"img",{"alt":36,"src":37},"cke_211.png","https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010117.36689244146635450773748801130604:20240906012345:2400:25FA9B8634564CDB776D9AE9C30845ACB19316CF89667922A243C3C7AB50771F.png",[],{"type":17,"tag":25,"props":40,"children":41},{},[42],{"type":17,"tag":34,"props":43,"children":46},{"alt":44,"src":45},"image.png","https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010125.86157506415307517398504228235116:20240906012345:2400:1F3DDCCD1718B914221E1032D6904F5E0D97AAEB49FDF8BD6DB485913F7A8316.png",[],{"type":17,"tag":25,"props":48,"children":49},{},[50],{"type":23,"value":51},"昇思MindSpore生态技术专家杨滔",{"type":17,"tag":25,"props":53,"children":54},{},[55],{"type":23,"value":56},"在为期四天的学习中，训练营从Transformer的基础知识入手，逐步探索了大模型技术的发展脉络。讲师们深入讲解了大模型的开发、预训练、微调和应用全流程的关键技术点，帮助大家建立了全面的技术全景图。通过对以BERT和GPT为代表的大模型的深入剖析，大家不仅理解了encoder-only和decoder-only结构的差异，还深入掌握了大模型训练过程中不可或缺的技术——分布式并行技术。此外，训练营还探讨了高效参数微调技术，提升大模型的训练效率，降低训练成本，并通过分析基于MindSpore框架的MusicGen和ChatPDF等大模型应用实例，让大家感受到大模型在现实中的无限可能。",{"type":17,"tag":25,"props":58,"children":59},{},[60],{"type":17,"tag":34,"props":61,"children":64},{"alt":62,"src":63},"cke_1542.png","https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010219.35389457053224949126644700607470:20240906012345:2400:FE05725440E6550E205AA6B693B5A701C18899B86EABB6915D3AFBF50253364A.png",[],{"type":17,"tag":25,"props":66,"children":67},{},[68],{"type":17,"tag":34,"props":69,"children":71},{"alt":44,"src":70},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010228.41167048888684692398180382525387:20240906012345:2400:E6849B45963219E748C83EE1180550EAC2C9C11E0C0A2B6E83DB32DAA2391AFB.png",[],{"type":17,"tag":25,"props":73,"children":74},{},[75],{"type":23,"value":76},"获奖同学合影",{"type":17,"tag":25,"props":78,"children":79},{},[80],{"type":23,"value":81},"以下仅为同学们在创新训练营中众多优秀作品中的部分璀璨亮点，它们不仅展现了同学们的智慧与才华，更是大家辛勤学习与不懈探索的结晶：",{"type":17,"tag":25,"props":83,"children":84},{},[85],{"type":17,"tag":86,"props":87,"children":88},"strong",{},[89,91,99],{"type":23,"value":90},"0",{"type":17,"tag":86,"props":92,"children":93},{},[94],{"type":17,"tag":86,"props":95,"children":96},{},[97],{"type":23,"value":98},"1",{"type":23,"value":100},"MindOverflow —— 针对线性代数的教学助手",{"type":17,"tag":25,"props":102,"children":103},{},[104],{"type":23,"value":105},"在大模型已经“飞入寻常百姓家”的时代，尽管通用预训练语言模型的能力日益增强，但在科研、教学与学习中，这些大模型的专业能力仍有待提升。尤其是在数学运算问题上的回答准确率，仍需进一步提高，以更好地满足高校教学的需求。",{"type":17,"tag":25,"props":107,"children":108},{},[109],{"type":23,"value":110},"为了解决这一问题，使大模型更好地服务于高校教学，MindOverflow团队利用MindSpore框架、RAG（Retrieval-Augmented Generation）算法以及Qwen-2预训练模型，基于高质量的线性代数数据集，开发了一款专业能力强大、交互界面完善的“AI助教”。这款“AI助教”能够准确回答线性代数问题，不仅有效减轻了教师的教学负担，还显著提升了学生的学习效果。",{"type":17,"tag":25,"props":112,"children":113},{},[114],{"type":17,"tag":34,"props":115,"children":117},{"alt":44,"src":116},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010246.12645278818392113254273571575062:20240906012345:2400:FAE8DD9437F2379CCA002C801AF843882C84C034C8706C38CC29EB70E1AFF604.png",[],{"type":17,"tag":25,"props":119,"children":120},{},[121,123],{"type":23,"value":122},"代码仓地址：",{"type":17,"tag":124,"props":125,"children":129},"a",{"href":126,"rel":127},"https://github.com/Lzzzzy4/Mind-Overflow",[128],"nofollow",[130],{"type":23,"value":126},{"type":17,"tag":25,"props":132,"children":133},{},[134],{"type":17,"tag":86,"props":135,"children":136},{},[137],{"type":23,"value":138},"02****三颗花椒队 —— image to text",{"type":17,"tag":25,"props":140,"children":141},{},[142],{"type":23,"value":143},"图生文技术旨在从图像中提取关键信息，并生成自然语言描述，从而实现对图像内容的深入理解和表达。然而，传统的图生文解决方案在训练过程中存在两个主要问题：首先，由于映射网络结构过于庞大，导致训练时间过长；其次，传统方案中所用的语言模型在处理中文文本时效果欠佳，导致生成的文本质量较差。",{"type":17,"tag":25,"props":145,"children":146},{},[147],{"type":23,"value":148},"为了解决这些问题，“三颗花椒”队的同学们提出了一系列改进措施。他们通过简化映射网络结构，显著缩短了模型的训练时间。同时，他们将传统的输出文本语言模型替换为中文处理能力更强的通义千问模型，以提升文本生成的效果和准确性。这些改进有效地提升了模型的训练性能，并改善了中文文本的生成效果。",{"type":17,"tag":25,"props":150,"children":151},{},[152],{"type":17,"tag":34,"props":153,"children":155},{"alt":44,"src":154},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010306.43292182765451506639428312506177:20240906012345:2400:CC39BBC84C4A95CD979506B7705CA6CF928AFAD145DF94A340ABC9303552A59E.png",[],{"type":17,"tag":25,"props":157,"children":158},{},[159],{"type":23,"value":160},"尽管取得了显著进展，但由于时间限制，当前生成的效果与预期仍存在一定差距。队员们计划在后续工作中，进一步优化数据集处理，精细调整映射网络并微调语言模型，选择文本生成能力更强的模型。此外，他们还考虑将映射层替换为表达能力更强的BERT等模型，以进一步提升整体效果。我们期待在不久的将来，能够看到他们更加完善的方案，并取得更加优秀的成果。",{"type":17,"tag":25,"props":162,"children":163},{},[164],{"type":23,"value":165},"效果展示：",{"type":17,"tag":25,"props":167,"children":168},{},[169],{"type":17,"tag":34,"props":170,"children":172},{"alt":44,"src":171},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010324.97134142730165795102160281356792:20240906012345:2400:B87E84FD2026CCB9F6979810F1458672953C3462746A5E518D2660F1E0CFFAED.png",[],{"type":17,"tag":25,"props":174,"children":175},{},[176],{"type":23,"value":122},{"type":17,"tag":25,"props":178,"children":179},{},[180],{"type":17,"tag":124,"props":181,"children":184},{"href":182,"rel":183},"https://xihe.mindspore.cn/projects/Devon/ImageToText",[128],[185],{"type":23,"value":182},{"type":17,"tag":25,"props":187,"children":188},{},[189],{"type":17,"tag":86,"props":190,"children":191},{},[192],{"type":23,"value":193},"03****大模型说的都队 —— 基于Yi-Chat的AI医疗诊断大师",{"type":17,"tag":25,"props":195,"children":196},{},[197],{"type":23,"value":198},"传统的医疗场景中，面临着人力成本高昂、城乡医疗服务质量差异明显等挑战。虽然大模型的引入在一定程度上提高了医疗服务的效率和水平，但其回答完全依赖于预训练的语言模型，这可能导致在处理复杂医学问题时出现不准确、不完整的回答，特别是由于缺乏最新的或专业的医学知识，可能导致严重的误导。",{"type":17,"tag":25,"props":200,"children":201},{},[202],{"type":23,"value":203},"针对这一问题，大模型说的都队的同学们深入研究了课程中讲解的高效参数微调技术和RAG算法，并成功开发了一个基于Yi-Chat的智能医疗问诊机器人。这个机器人通过首先检索与特定病症相关的最新和专业信息，然后再让大模型基于这些检索到的信息生成回答，从而有效解决了大模型在医疗场景中常见的“幻觉”问题。这样的技术方案，不仅显著提升了问诊机器人的准确性，还大大增强了其在复杂医学问题处理上的可靠性，展示了学员们在本次学习活动中的创新能力和技术深度。",{"type":17,"tag":25,"props":205,"children":206},{},[207],{"type":17,"tag":34,"props":208,"children":210},{"alt":44,"src":209},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010343.63840098240019865426412092484722:20240906012345:2400:370BDB72307A403E4B9C6A324D9DB3FEDC039ABE9B6D260C6598D532A883566E.png",[],{"type":17,"tag":25,"props":212,"children":213},{},[214,219,221],{"type":17,"tag":86,"props":215,"children":216},{},[217],{"type":23,"value":218},"04****baseline",{"type":23,"value":220}," ",{"type":17,"tag":86,"props":222,"children":223},{},[224],{"type":23,"value":225},"—— AI专用搜索增强",{"type":17,"tag":25,"props":227,"children":228},{},[229],{"type":23,"value":230},"在信息爆炸的时代洪流中，众多网站深藏信息丰富的宝藏帖，却受限于低效精准的关键词搜索。传统的方法普遍存在效率低下、精准度不足以及广告泛滥等痛点问题，baseline队旨在利用大模型为这些网站赋能，通过一系列创新性的实现方案，旨在为用户带来便捷舒适的搜索体验：",{"type":17,"tag":25,"props":232,"children":233},{},[234],{"type":23,"value":235},"智能提取关键词生成搜索请求：GLM+Prompting技术深度解析用户输入，精准提取并生成高效搜索请求，提升初始准确性。",{"type":17,"tag":25,"props":237,"children":238},{},[239],{"type":23,"value":240},"精准捕获答案构建知识库：RAG策略筛选整合高质量答案，确保信息丰富且有效，避免无效信息干扰。",{"type":17,"tag":25,"props":242,"children":243},{},[244],{"type":23,"value":245},"AI整合输出最优答案：Bert+Fine-Tuning技术深度匹配问题与答案，自动化处理呈现快速准确结果。",{"type":17,"tag":25,"props":247,"children":248},{},[249],{"type":23,"value":250},"昇思MindSpore框架加速大模型部署：提升上下文理解与开发效率，使搜索结果更贴近用户真实意图。",{"type":17,"tag":25,"props":252,"children":253},{},[254],{"type":17,"tag":34,"props":255,"children":257},{"alt":44,"src":256},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010359.71341103893428339755685020959201:20240906012345:2400:91EA5391B18A46E403F4EA541A5E6C847D694670924AF2F2FAAF8BF37B786FDB.png",[],{"type":17,"tag":25,"props":259,"children":260},{},[261],{"type":23,"value":165},{"type":17,"tag":25,"props":263,"children":264},{},[265],{"type":17,"tag":34,"props":266,"children":268},{"alt":44,"src":267},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010417.68794012030899137706927009412976:20240906012345:2400:8931C3101E6B78A3C7B1FD9E53FB5C08B16C590EA8719770B627ED8D38A95D0D.png",[],{"type":17,"tag":25,"props":270,"children":271},{},[272],{"type":17,"tag":34,"props":273,"children":275},{"alt":44,"src":274},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010429.42568132755226578567575431876382:20240906012345:2400:A2C7F53FED76646B5406DB97DCA3DF26EF457F28BA331E4E03587215ADC190EE.png",[],{"type":17,"tag":25,"props":277,"children":278},{},[279],{"type":17,"tag":86,"props":280,"children":281},{},[282],{"type":23,"value":283},"05****我写的都队 —— 基于bert真假新闻鉴别器及基于gpt的新闻扩写器",{"type":17,"tag":25,"props":285,"children":286},{},[287],{"type":23,"value":288},"我写的都对队的同学们之前并不了解深度学习和大模型方面的知识，在课上接触和了解到了一些NLP以及大模型的知识后，想要实现一个新闻的鉴别和生成课题中。他们灵活地将课堂上讲到的bert的文本分类GPT的文本生成的例子运用到了大作业课题中。同学们分别将两个模型作为文本分类器和扩写器，实现了新闻的分类和生成并取得了良好的效果。他们还联想到了使用GAN网络的生成器和判别器进行对抗训练，强化作品生成的能力和准确性，他们将会在后续参与到开源实习活动中，尝试实现对课题的进一步优化。",{"type":17,"tag":25,"props":290,"children":291},{},[292],{"type":23,"value":165},{"type":17,"tag":25,"props":294,"children":295},{},[296],{"type":17,"tag":34,"props":297,"children":299},{"alt":44,"src":298},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010448.72198835861604412371396297365447:20240906012345:2400:5191C199F0297FE0D1B330DDEB1F2E09E4C8946E43FDAA9A1ADE0D0BDECA3C57.png",[],{"type":17,"tag":25,"props":301,"children":302},{},[303,304],{"type":23,"value":122},{"type":17,"tag":124,"props":305,"children":308},{"href":306,"rel":307},"https://xihe.mindspore.cn/projects/lucy/project/",[128],[309],{"type":23,"value":306},{"type":17,"tag":25,"props":311,"children":312},{},[313],{"type":23,"value":314},"接下来，让我们一同聆听同学们的学习心得，感受他们在LLM领域探索旅程中的收获与成长。",{"type":17,"tag":25,"props":316,"children":317},{},[318],{"type":23,"value":319},"“我只有一些深度学习基础，原来有些担心跟不上。后来我发现完全能跟上老师们的节奏，…，感觉老师们不急于介绍MindSpore的细节，而是很细心地从基础讲起，然后很自然地用MindSpore实现——这真的要点一个大大的赞！”",{"type":17,"tag":25,"props":321,"children":322},{},[323],{"type":23,"value":324},"“在这三天的课程中，我印象最深的是分布式并行技术的讲解，从数据拆分，模型拆分，再到pipeline的层级拆分，每一种新的方式和思路的出现，与他们所面临的不同的问题和代价，一次次的拓宽着我的思路…”",{"type":17,"tag":25,"props":326,"children":327},{},[328],{"type":23,"value":329},"“收获最大的部分就是最后做项目的时候，用一个下午和晚上从0到1训了个模型，并不断的尝试扩展其功能（前端、目标检测等），体验了一夜一个奇迹的快乐”",{"type":17,"tag":25,"props":331,"children":332},{},[333],{"type":23,"value":334},"“我印象最深的部分是GPT1与GPT2的学习。在日常生活的很多方面，我都会用到GPT，但作为大模型方面的小白，我一直对GPT背后的原理了解甚少。通过学习这次课程，我了解了GPT等大模型的原理，并且尝试了大模型下游应用的编写，让我对大模型有了更深的了解。”",{"type":17,"tag":25,"props":336,"children":337},{},[338],{"type":23,"value":339},"“印象最深的是基于Transformer的机器翻译实操。自己一步步构建出这一复杂的模型并运行起来，看到良好的效果是一件神奇而有趣的事情。”",{"type":17,"tag":25,"props":341,"children":342},{},[343],{"type":23,"value":344},"“参加昇思MindSpore训练营让我深刻感受到了人工智能的魅力与潜力。课程中理论与实践的结合，使我不仅在技术上有所提升，更在思维上得到了启发。通过实际操作，我掌握了深度学习和自然语言处理等核心技术，并在项目中应用这些技术解决了实际问题。我意识到， AI不仅是技术的革新，更是对人类创造力的一种拓展。在这几天和其他学校队友的朝夕相处，合作攻克难题，这样宝贵的经历也让我们产生了更深的友谊。”",{"type":17,"tag":25,"props":346,"children":347},{},[348],{"type":23,"value":349},"希望本次创新训练营能够作为同学们在知识海洋中航行的起点，祝愿每位同学在未来的大模型领域里继续扬帆起航，创造更多辉煌成就。",{"type":17,"tag":25,"props":351,"children":352},{},[353],{"type":23,"value":354},"在最后的结营仪式上，昇思MindSpore AI框架业务副总经理孙宾表示：“昇思MindSpore AI框架作为软件根技术，已成为加速人工智能、大模型开发、推动产业发展的核心力量，广泛应用于运营商、金融、交通、互联网等行业。昇思MindSpore自开源以来一直重视人才的培养，人才是AI生态的源头活水，目前已发布了四项行动，持续赋能AI人才培育，包括学术论文基金、开发板应用创新、原生大模型孵化以及开源社区实习活动等。”最后，孙宾鼓励学生们继续参与昇思MindSpore社区活动，共同推动人工智能技术的发展。",{"type":17,"tag":25,"props":356,"children":357},{},[358],{"type":17,"tag":34,"props":359,"children":361},{"alt":44,"src":360},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010507.00497142061365443987166230554141:20240906012345:2400:A21E9ED0F57937B57CA01E177233AA2602CFF02693A858CFD0F76DF806428819.png",[],{"type":17,"tag":25,"props":363,"children":364},{},[365],{"type":23,"value":366},"昇思MindSpore AI框架业务副总经理孙宾",{"type":17,"tag":25,"props":368,"children":369},{},[370],{"type":23,"value":371},"ICT销服招聘调配部部长傅丹表示华为一直致力于推动数字化进程，为全球用户提供更便捷、高效的服务。同时，她也提到，华为高度重视人才的培养和发展，拥有完善的人才机制和培养体系，为员工提供广阔的发展空间和机会。最后，傅丹再次表达了对与会同学的热烈欢迎，并鼓励大家积极了解华为，共同探索未来的发展机会。",{"type":17,"tag":25,"props":373,"children":374},{},[375],{"type":17,"tag":34,"props":376,"children":378},{"alt":44,"src":377},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010545.53751546642365909456702073779885:20240906012345:2400:130107093E7B00C2088A0DBA0857B2216D001A99276B89116A262299378076FC.png",[],{"type":17,"tag":25,"props":380,"children":381},{},[382],{"type":23,"value":383},"ICT销服招聘调配部部长傅丹",{"type":17,"tag":25,"props":385,"children":386},{},[387],{"type":23,"value":388},"最后，上海交通大学学生创新中心书记熊振华对本次的创新训练营进行了总结，他指出，今年的暑期训练营虽然只有短短的四天，但却过得充实且愉快，各个高校的同学们在上海交大结下了深厚的友谊，相信这段经历都将成为同学们青春岁月中难以忘怀的记忆。",{"type":17,"tag":25,"props":390,"children":391},{},[392],{"type":17,"tag":34,"props":393,"children":395},{"alt":44,"src":394},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20240906010603.18922681968196089549442414946475:20240906012345:2400:6F2919C01501E5E9BE574A4C40805386AE73BEC6DE2EAA5E4D23940445D145B6.png",[],{"type":17,"tag":25,"props":397,"children":398},{},[399],{"type":23,"value":400},"上海交通大学学生创新中心书记熊振华",{"type":17,"tag":25,"props":402,"children":403},{},[404],{"type":23,"value":405},"长三角研究型高校联盟的创新训练营已迎来了第五年的成功举办，吸引了更多学校和学生的参与，不仅促进了优秀学生之间的交流，还为他们将来参加华为ICT大赛及昇思MindSpore社区开源实习提供了坚实的基础。未来，昇思MindSpore开源社区将携手更多高校，不断地培养人工智能领域发展需要的创新型&应用型人才。",{"title":7,"searchDepth":407,"depth":407,"links":408},4,[],"markdown","content:news:zh:3378.md","content","news/zh/3378.md","news/zh/3378","md",1776506083218]