[{"data":1,"prerenderedAt":329},["ShallowReactive",2],{"content-query-Sm2VFbwwAY":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"body":13,"_type":323,"_id":324,"_source":325,"_file":326,"_stem":327,"_extension":328},"/news/zh/3785","zh",false,"","0day同步！昇思MindSpore首发支持智谱GLM-4.1V-9B-Thinking多模态模型","7月2日，智谱AI发布视觉语言大模型GLM-4.1V-9B-Thinking，昇思MindSpore在开源当日即实现0day支持。","2025-07-02","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/07/04/2786abd0c3bc46829d106d73bae78e51.png","news",{"type":14,"children":15,"toc":320},"root",[16,24,30,35,46,51,60,72,83,88,93,101,112,137,145,156,161,171,179,190,195,203,208,216,221,229,234,242,250,261,266,271,276,284,289,297,309],{"type":17,"tag":18,"props":19,"children":21},"element","h1",{"id":20},"_0day同步昇思mindspore首发支持智谱glm-41v-9b-thinking多模态模型",[22],{"type":23,"value":8},"text",{"type":17,"tag":25,"props":26,"children":27},"p",{},[28],{"type":23,"value":29},"7月2日，智谱AI发布视觉语言大模型GLM-4.1V-9B-Thinking，昇思MindSpore在开源当日即实现0day支持。昇思MindSpore通过提供主流生态接口，依托MindSpore One套件，小时级完成模型无缝迁移，并将版本上传至开源社区，欢迎广大开发者下载体验！",{"type":17,"tag":25,"props":31,"children":32},{},[33],{"type":23,"value":34},"MindSpore One代码仓：",{"type":17,"tag":25,"props":36,"children":37},{},[38],{"type":17,"tag":39,"props":40,"children":44},"a",{"href":41,"rel":42},"https://github.com/mindspore-lab/mindone",[43],"nofollow",[45],{"type":23,"value":41},{"type":17,"tag":25,"props":47,"children":48},{},[49],{"type":23,"value":50},"体验指导链接：",{"type":17,"tag":25,"props":52,"children":53},{},[54],{"type":17,"tag":39,"props":55,"children":58},{"href":56,"rel":57},"https://modelers.cn/models/MindSpore-Lab/GLM-4.1V-9B-Thinking",[43],[59],{"type":23,"value":56},{"type":17,"tag":25,"props":61,"children":62},{},[63],{"type":17,"tag":64,"props":65,"children":66},"strong",{},[67],{"type":17,"tag":64,"props":68,"children":69},{},[70],{"type":23,"value":71},"# 01",{"type":17,"tag":25,"props":73,"children":74},{},[75],{"type":17,"tag":64,"props":76,"children":77},{},[78],{"type":17,"tag":64,"props":79,"children":80},{},[81],{"type":23,"value":82},"模型介绍",{"type":17,"tag":25,"props":84,"children":85},{},[86],{"type":23,"value":87},"视觉语言大模型（VLM）已经成为智能系统的关键基石。随着真实世界的智能任务越来越复杂，VLM模型也亟需在基本的多模态感知之外， 逐渐增强复杂任务中的推理能力，提升自身的准确性、全面性和智能化程度，使得复杂问题解决、长上下文理解、多模态智能体等智能任务成为可能。",{"type":17,"tag":25,"props":89,"children":90},{},[91],{"type":23,"value":92},"基于 GLM-4-9B-0414 基座模型，新版VLM开源模型 GLM-4.1V-9B-Thinking 引入思考范式，通过课程采样强化学习 RLCS（Reinforcement Learning with Curriculum Sampling）全面提升模型能力， 达到 10B 参数级别的视觉语言模型的最强性能，在18个榜单任务中持平甚至超过8倍参数量的 Qwen-2.5-VL-72B。",{"type":17,"tag":25,"props":94,"children":95},{},[96],{"type":17,"tag":64,"props":97,"children":98},{},[99],{"type":23,"value":100},"# 02",{"type":17,"tag":25,"props":102,"children":103},{},[104],{"type":17,"tag":64,"props":105,"children":106},{},[107],{"type":17,"tag":64,"props":108,"children":109},{},[110],{"type":23,"value":111},"环境要求",{"type":17,"tag":113,"props":114,"children":115},"ul",{},[116,122,127,132],{"type":17,"tag":117,"props":118,"children":119},"li",{},[120],{"type":23,"value":121},"mindspore 2.5.0/2.6.0",{"type":17,"tag":117,"props":123,"children":124},{},[125],{"type":23,"value":126},"ascend driver 24.1.RC3.b080",{"type":17,"tag":117,"props":128,"children":129},{},[130],{"type":23,"value":131},"firmware 7.5.T11.0.B088",{"type":17,"tag":117,"props":133,"children":134},{},[135],{"type":23,"value":136},"cann tookit/kernel 8.0.RC3.beta1",{"type":17,"tag":25,"props":138,"children":139},{},[140],{"type":17,"tag":64,"props":141,"children":142},{},[143],{"type":23,"value":144},"# 03",{"type":17,"tag":25,"props":146,"children":147},{},[148],{"type":17,"tag":64,"props":149,"children":150},{},[151],{"type":17,"tag":64,"props":152,"children":153},{},[154],{"type":23,"value":155},"权重下载",{"type":17,"tag":25,"props":157,"children":158},{},[159],{"type":23,"value":160},"下载权重到/home/GLM-4.1V，需要预留至少21GB空间，可以根据需要调整路径。",{"type":17,"tag":162,"props":163,"children":165},"pre",{"code":164},"\npip install openmind_hub\nexport HUB_WHITE_LIST_PATHS=/home/GLM-4.1V\npython\nfrom openmind_hub import snapshot_download\nsnapshot_download(\n    repo_id=\"MindSpore-Lab/GLM-4.1V-9B-Thinking\",\n    local_dir=\"/home/GLM-4.1V\",\n    local_dir_use_symlinks=False\n)\n",[166],{"type":17,"tag":167,"props":168,"children":169},"code",{"__ignoreMap":7},[170],{"type":23,"value":164},{"type":17,"tag":25,"props":172,"children":173},{},[174],{"type":17,"tag":64,"props":175,"children":176},{},[177],{"type":23,"value":178},"# 04",{"type":17,"tag":25,"props":180,"children":181},{},[182],{"type":17,"tag":64,"props":183,"children":184},{},[185],{"type":17,"tag":64,"props":186,"children":187},{},[188],{"type":23,"value":189},"环境准备",{"type":17,"tag":25,"props":191,"children":192},{},[193],{"type":23,"value":194},"安装最新的MinSpore One以及Transformers，/home路径根据需要调整。",{"type":17,"tag":162,"props":196,"children":198},{"code":197},"\ncd /home\ngit clone https://github.com/mindspore-lab/mindone.git --depth 1\n",[199],{"type":17,"tag":167,"props":200,"children":201},{"__ignoreMap":7},[202],{"type":23,"value":197},{"type":17,"tag":25,"props":204,"children":205},{},[206],{"type":23,"value":207},"若执行失败，可以尝试：",{"type":17,"tag":162,"props":209,"children":211},{"code":210},"\ngit clone https://gh-proxy.com/github.com/mindspore-lab/mindone.git --depth 1\n",[212],{"type":17,"tag":167,"props":213,"children":214},{"__ignoreMap":7},[215],{"type":23,"value":210},{"type":17,"tag":25,"props":217,"children":218},{},[219],{"type":23,"value":220},"之后执行：",{"type":17,"tag":162,"props":222,"children":224},{"code":223},"cd mindone\npip install -e .\n",[225],{"type":17,"tag":167,"props":226,"children":227},{"__ignoreMap":7},[228],{"type":23,"value":223},{"type":17,"tag":25,"props":230,"children":231},{},[232],{"type":23,"value":233},"安装transformers（安装过程中忽略报错信息）：",{"type":17,"tag":162,"props":235,"children":237},{"code":236},"\npip install transformers==4.53.0\n",[238],{"type":17,"tag":167,"props":239,"children":240},{"__ignoreMap":7},[241],{"type":23,"value":236},{"type":17,"tag":25,"props":243,"children":244},{},[245],{"type":17,"tag":64,"props":246,"children":247},{},[248],{"type":23,"value":249},"# 05",{"type":17,"tag":25,"props":251,"children":252},{},[253],{"type":17,"tag":64,"props":254,"children":255},{},[256],{"type":17,"tag":64,"props":257,"children":258},{},[259],{"type":23,"value":260},"执行推理测试",{"type":17,"tag":25,"props":262,"children":263},{},[264],{"type":23,"value":265},"参数model_name指的是本地下载的权重，根据需要调整。image指的是本地存储的图片，需要更改。",{"type":17,"tag":25,"props":267,"children":268},{},[269],{"type":23,"value":270},"默认推理token数是128，如果需要更多可以修改/home/mindone/examples/transformers/glm4v/generate.py中的max_new_tokens=128改成2048。",{"type":17,"tag":25,"props":272,"children":273},{},[274],{"type":23,"value":275},"若想要输出英文，则输入英文（推荐）",{"type":17,"tag":162,"props":277,"children":279},{"code":278},"cd /home/mindone/examples/transformers/glm4v\npython generate.py --model_name /home/GLM-4.1V --image /home/trial.jpg --prompt \"Describe this image.\"\n",[280],{"type":17,"tag":167,"props":281,"children":282},{"__ignoreMap":7},[283],{"type":23,"value":278},{"type":17,"tag":25,"props":285,"children":286},{},[287],{"type":23,"value":288},"若想要输出中文，则输入中文：",{"type":17,"tag":162,"props":290,"children":292},{"code":291},"cd /home/mindone/examples/transformers/glm4v\npython generate.py --model_name /home/GLM-4.1V --image /home/trial.jpg --prompt \"请帮我描述这张图片。\"\n",[293],{"type":17,"tag":167,"props":294,"children":295},{"__ignoreMap":7},[296],{"type":23,"value":291},{"type":17,"tag":25,"props":298,"children":299},{},[300,302],{"type":23,"value":301},"昇思MindSpore AI框架将持续支持更多业界主流模型，欢迎大家试用并在昇思论坛提供宝贵的反馈意见。（",{"type":17,"tag":39,"props":303,"children":306},{"href":304,"rel":305},"https://discuss.mindspore.cn/%EF%BC%89",[43],[307],{"type":23,"value":308},"https://discuss.mindspore.cn/）",{"type":17,"tag":25,"props":310,"children":311},{},[312,314],{"type":23,"value":313},"MindSpore Transformers SIG首次月例会将在7月3日16：00举办，欢迎感兴趣的开发者参与：",{"type":17,"tag":39,"props":315,"children":318},{"href":316,"rel":317},"https://meeting.tencent.com/dm/olkrpDKfadmJ",[43],[319],{"type":23,"value":316},{"title":7,"searchDepth":321,"depth":321,"links":322},4,[],"markdown","content:news:zh:3785.md","content","news/zh/3785.md","news/zh/3785","md",1776506089298]