[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"site-home-live-entry-setting":3,"site-header-contests":8,"site-header-camps":49,"study-material-detail-study-27":104},{"enabled":4,"title":5,"statusText":5,"buttonText":5,"campSlug":5,"stage":6,"classroomId":7,"hasClassroom":4},false,"",0,null,{"total":9,"list":10,"pageNum":44,"pageSize":45,"size":9,"startRow":44,"endRow":9,"pages":44,"prePage":6,"nextPage":6,"isFirstPage":46,"isLastPage":46,"hasPreviousPage":4,"hasNextPage":4,"navigatePages":47,"navigatepageNums":48,"navigateFirstPage":44,"navigateLastPage":44},3,[11,25,35],{"slug":12,"title":13,"logo":7,"liveUrl":14,"prize":15,"sponsor":16,"description":17,"organizer":18,"startTime":19,"endTime":20,"registrationDeadline":21,"status":22,"dashboardVisibility":23,"createdAt":24},"spring2026","2026春季启元人工智能大赛","https:\u002F\u002Fwww.bilibili.com\u002Fblackboard\u002Flive\u002Flive-activity-player.html?cid=1822703816&quality=0","200,000","麒麟软件有限公司","“2026春季启元人工智能大赛”是由启元实验室主办的面向全国人工智能领域开发者、科研人员及高校学生的权威赛事。大赛旨在促进与培养具备工程实践能力和创新能力的优秀人工智能领域开发者、推广国产软硬件技术与生态、助力国家人工智能产业高质量发展。\n各参赛选手需在规定的比赛周期内，依据赛事规则选定赛题并完成相应挑战，优胜者将共同角逐总额20万元的赛事奖金池，并且还将有机会获得实习和知名企业优质岗位内推资格。","启元实验室","2026-04-19 16:00:00","2026-07-13 16:00:00","2026-06-16 16:00:00","ongoing","private","2026-04-16 08:00:00",{"slug":26,"title":27,"logo":7,"liveUrl":14,"prize":28,"sponsor":7,"description":29,"organizer":18,"startTime":30,"endTime":31,"registrationDeadline":32,"status":33,"dashboardVisibility":23,"createdAt":34},"autumn2025","2025秋季启元人工智能大赛","150,000","“2025秋季启元人工智能大赛”是由启元实验室主办的面向全国人工智能领域开发者、科研人员及高校学生的权威赛事。大赛旨在促进与培养具备工程实践能力和创新能力的优秀人工智能领域开发者、推广国产软硬件技术与生态、助力国家人工智能产业高质量发展。","2025-11-08 08:00:00","2026-01-13 08:00:00","2025-12-30 08:00:00","ended","2025-11-03 08:00:00",{"slug":36,"title":37,"logo":7,"liveUrl":14,"prize":38,"sponsor":7,"description":39,"organizer":18,"startTime":40,"endTime":41,"registrationDeadline":42,"status":33,"dashboardVisibility":23,"createdAt":43},"summer2025","启元人工智能大赛(系统方向)","80,000","启元人工智能大赛由启元实验室主办，其中系统赛道基于启元实验室九源软件栈的相关开源项目，通过拉动开源社区力量，一方面促进国产智能计算生态的发展，另一方面也为人工智能系统方向的学生及从业者提供一个实践平台。","2025-07-08 08:00:00","2025-08-25 08:00:00","2025-07-22 08:00:00","2025-06-22 08:00:00",1,5,true,8,[44],{"total":45,"list":50,"pageNum":44,"pageSize":45,"size":45,"startRow":44,"endRow":45,"pages":44,"prePage":6,"nextPage":6,"isFirstPage":46,"isLastPage":46,"hasPreviousPage":4,"hasNextPage":4,"navigatePages":47,"navigatepageNums":103,"navigateFirstPage":44,"navigateLastPage":44},[51,62,73,84,94],{"id":52,"slug":53,"title":54,"logo":55,"publicLiveEmbedUrl":14,"description":56,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":16,"startTime":58,"endTime":59,"status":22,"dashboardVisibility":60,"createdAt":58,"updatedAt":61},"019eeeff-6b00-7d55-a629-efafded3e232","summer2026","2026年夏季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fuploads\u002F2026-06\u002F019eef7f-7816-7828-9309-79c2c61dd568.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=uiV9kOffa9Xq8DT%2BbErZU0JyQ98%3D","由启元实验室发起的InfiniTensor大模型与人工智能系统训练营，面向开源社区、高校及产业界伙伴开放。活动以「普及AI全栈知识技术、培养工程实践与创新型人才、推广国产软硬件技术生态」为核心目标，助力国家人工智能产业高质量发展。","...","2026-06-22 08:00:00","2026-10-19 08:00:00","public","2026-10-12 08:00:00",{"id":63,"slug":64,"title":65,"logo":66,"publicLiveEmbedUrl":7,"description":67,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":68,"startTime":69,"endTime":70,"status":33,"dashboardVisibility":23,"createdAt":71,"updatedAt":72},"019af3f3-e06a-7245-aa8d-37dc99a69c73","winter2025","2025年冬季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fuploads\u002F2025-12\u002F019afc69-b7ca-7805-91fb-605429671c01.png?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=8u%2BaQaeliN4QzF6MZWiB4jHYDCE%3D","由启元实验室发起的InfiniTensor大模型与人工智能系统训练营，面向开源社区、高校及产业界伙伴开放。活动以「普及AI全栈知识技术、培养工程实践与创新型人才、推广国产软硬件技术生态」为核心目标，助力国家人工智能产业高质量发展。 \n","新华三技术有限公司","2025-12-25 08:00:00","2026-03-27 08:00:00","2025-12-06 08:00:00","2026-10-22 20:24:30",{"id":74,"slug":36,"title":75,"logo":76,"publicLiveEmbedUrl":7,"description":77,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":78,"startTime":79,"endTime":80,"status":81,"dashboardVisibility":23,"createdAt":82,"updatedAt":83},"01980342-208c-7803-b3f3-19e1d31960ba","2025年夏季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fuploads\u002F2025-07\u002F0198045e-28fe-78f0-95fe-23835ce58395.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=8pXXbnSlNEtVJCJ1e4RzUlqmRn8%3D","InfiniTensor大模型与人工智能系统训练营由启元实验室及清华大学 AI 系统方向团队发起，面向开源社区、高校和产业界，普及人工智能各方面知识和技术，培养相关人才。训练营联合开源项目开发者、学术界大牛、工业界资深工程师参与授课与项目指导。课程采用「线上直播授课 + 在线答疑 + 项目实践辅导 」的方式进行，确保每位同学都能获得充分成长。训练营全程免费，通过视频、直播课、习题和项目实践，在人工智能飞速发展的背景下，为学员提供一窥深层技术的机会。","北京积算科技有限公司","2025-07-14 08:00:00","2025-09-30 08:00:00","upcoming","2025-07-13 08:00:00","2026-10-22 20:24:58",{"id":85,"slug":86,"title":87,"logo":88,"publicLiveEmbedUrl":7,"description":89,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":5,"startTime":90,"endTime":91,"status":81,"dashboardVisibility":23,"createdAt":92,"updatedAt":93},"01980cf9-a7d8-7796-ae86-0bbe05334add","winter2024","2024年冬季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Ftesting-cdn.infinitensor.com\u002Fuploads\u002F2025-07\u002F01981269-ce23-7624-a287-9f07dd0d85f6.jpg","InfiniTensor大模型与人工智能系统训练营由清华大学发起，面向开源社区、高校和产业界，普及人工智能各方面知识和技术，培养相关人才。训练营全程免费，通过视频、直播课、习题和项目实践，在人工智能飞速发展的背景下，为学员提供一窥深层技术的机会。报名学员请阅读“导学”课程页的训练营介绍。","2025-01-05 08:00:00","2025-03-22 08:00:00","2025-07-15 08:00:00","2026-10-22 20:24:39",{"id":95,"slug":96,"title":97,"logo":98,"publicLiveEmbedUrl":7,"description":99,"introduction":57,"organizer":18,"coOrganizer":5,"sponsor":5,"startTime":100,"endTime":101,"status":81,"dashboardVisibility":23,"createdAt":92,"updatedAt":102},"01980c20-363a-7dc3-8f5a-58a68e08dcc4","summer2024","2024年夏季InfiniTensor大模型与人工智能系统训练营","https:\u002F\u002Ftesting-cdn.infinitensor.com\u002Fuploads\u002F2025-07\u002F01981269-fbfa-788a-afcd-17fa9be74ff9.jpg","InfiniTensor大模型与人工智能系统训练营由清华大学发起，面向开源社区和高校，普及人工智能各方面知识和技术，培养相关人才。训练营全程免费，通过课程、习题和项目实践，在人工智能飞速发展的背景下，为学员提供一窥深层技术的机会。","2024-07-20 08:00:00","2024-09-29 08:00:00","2026-10-22 20:24:49",[44],{"currentItem":105,"processedData":125},{"img":106,"keywords":107,"treeType":44,"description":108,"showStatus":44,"updateTime":109,"type":110,"title":111,"content":112,"nextArticle":113,"createTime":109,"sortOrder":117,"canonicalPath":118,"id":119,"contentType":120,"prevArticle":121},"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_fm.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=%2F1Cc6FaLpdePxA6xtPbamt1J34U%3D","InfiniTensor,冬季训练营,AI编译器,PyTorch,图优化","聚焦 PyTorch 图优化方法，解析计算图层面的性能优化路径与编译协同思路。","2026-10-22 20:38:11",2,"InfiniTensor 冬季训练营 AI 编译器第三课《PyTorch 中图优化》","![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw01.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=DaFm%2BE4%2BgzpflYTWcPl1GVVejWI%3D)\r\n\r\n\r\n2026 年 1 月 20 日，2025 冬季 InfiniTensor 大模型与人工智能系统训练营 AI 编译器系列第三课《PyTorch 中图优化》开讲。\r\n\r\n本节课主要讲解 PyTorch 核心设计理念与图优化技术体系演进。\r\n\r\n### PyTorch 基础介绍\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw02.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=sKq%2BoLjfVo3BVtTFKGHXQpImO24%3D)\r\n\r\n\r\n### 动态计算图的优势与劣势\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw03.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=EIxakjFg2tfQ2ugzuZKMvaBoJWE%3D)\r\n\r\n\r\n### PyTorch 中的图优化\r\n- 前端：在保持 Python 动态语义的前提下，对张量相关计算进行捕获与抽象，生成可优化的中间表示（IR）\r\n- 后端：基于中间表示进行图级优化、调度与代码生成\r\n\r\n#### 1. TorchScript\r\n- 介绍：通过`torch.jit`实现\r\n- 两个接口：\r\n  - `torch.jit.trace`\r\n  - `torch.jit.script`\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw04.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=lkEGz8fjErj1EWPdsq37V0gBUFo%3D)\r\n\r\n\r\n#### 2. torch.fx\r\n- 介绍：Python 层 IR 与图变换工具\r\n- 技术亮点：\r\n  - **Symbolic Tracing**：通过 Proxy 对象代替输入张量，记录所有张量运算的结构与路径。\r\n\r\n  - **fx IR**：高度简洁，6 种核心指令，`placeholder`、`get_attr`、`call_function`、`call_module`、`call_method`、`output`。\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw05.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=IUHD5QHYKQTBhy3nk2WjM%2BGjjGA%3D)\r\n\r\n\r\n  - **Python 源到源转换**：\r\n     - IR 反向映射回 Python 源码\r\n     - 基于 FX Graph 修改模型结构，导出新的`nn.Module`代码\r\n\r\n#### 3. PyTorch 2.0 时代：torch.compile\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw06.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=gwK1wQapAQ8%2F7OYBBgBoSFt7Th8%3D)\r\n\r\n- 介绍：统一编译接口，在尽量保留 Python 动态语义的前提下，通过图捕获与优化提升执行效率\r\n- 组件与步骤：\r\n  - **TorchDynamo（前端）**：解释器级 tracing 框架，在 Python 字节码层面捕获张量相关计算，生成受 guard 约束的 FX Graph\r\n    ![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw07.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=3%2BNpWAxNPJQzFPHSl0TjyF3TZD0%3D)\r\n    - 工作原理：拦截 Python 函数执行，记录张量相关 IR\r\n    - Guards 机制：监控变量类型与值，确保重用编译图\r\n    - Graph Break：子图划分，处理无法转换为静态图的部分\r\n\r\n\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw08.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=gp9DY2lub3I3VzRChN47wLt%2BTaM%3D)\r\n\r\n\r\n  - **AOT Autograd**：提前生成前向与反向联合计算图\r\n    - 工作流程：Dispatcher 机制拦截与重定义 Tensor 操作，切分正反向计算图\r\n\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw09.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=7bOsjtz9tJ1WFto9F%2Bd%2FTlNRw1c%3D)\r\n\r\n  - **TorchInductor（后端）**：后端优化，生成高效可执行代码\r\n    - 工作流程：接受 FX Graph，进行算子融合、内存规划与调度优化，并生成高效的执行代码\r\n    - 示例：CUDA 设备上的 Triton 后端代码生成过程\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw10.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=mw0rbM9MpYqFP1QfgFJggnZJm3w%3D)\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw11.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=BZNZftc4LH5FHqAQxAvlyHqIT9g%3D)\r\n\r\n\r\n### 总结\r\n本节课讲了 PyTorch 图优化技术体系的演进历程与核心特性。后续课程将讲解 AI 编译器中的后端优化。\r\n\r\n\r\n\r\n\r\n![2025冬季训练营课程表](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_01_zw6.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=DEth0OhERyx68xufLNiv%2BIt7foE%3D)\r\n\r\n### 💡 学习贴士\r\n更多细节及完整课程内容，请观看直播或查看课程回放。\r\n\r\n链接：https:\u002F\u002Fwww.infinitensor.com\u002Fcamp\u002Fwinter2025\r\n\r\n### 📚 训练营报名持续进行中\r\n扫码报名与加入群聊，或点击 **「阅读原文」** 直达官网报名\r\n\r\n\r\n![](https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_01_zw7.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=OXZH4IYHRMHbFPG7EYCBYHhSMto%3D)",{"publishTime":7,"createTime":109,"canonicalPath":114,"id":115,"title":116},"\u002Fstudy-material\u002Fstudy-26","study-26","InfiniTensor 冬季训练营训练系统第二课《分布式训练》",82,"\u002Fstudy-material\u002Fstudy-27","study-27","markdown",{"publishTime":7,"createTime":109,"canonicalPath":122,"id":123,"title":124},"\u002Fstudy-material\u002Fstudy-28","study-28","InfiniTensor 冬季训练营推理引擎第三课《从大模型推理到 AI 对话》",{"htmlContent":126,"headings":127},"\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw01.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=DaFm%2BE4%2BgzpflYTWcPl1GVVejWI%3D\" alt=\"\">\u003C\u002Fp>\n\u003Cp>2026 年 1 月 20 日，2025 冬季 InfiniTensor 大模型与人工智能系统训练营 AI 编译器系列第三课《PyTorch 中图优化》开讲。\u003C\u002Fp>\n\u003Cp>本节课主要讲解 PyTorch 核心设计理念与图优化技术体系演进。\u003C\u002Fp>\n\u003Ch3 id=\"pytorch-基础介绍\">\u003Ca class=\"anchor\" href=\"#pytorch-基础介绍\" aria-label=\"跳转到标题：PyTorch 基础介绍\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>PyTorch 基础介绍\u003C\u002Fh3>\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw02.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=sKq%2BoLjfVo3BVtTFKGHXQpImO24%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch3 id=\"动态计算图的优势与劣势\">\u003Ca class=\"anchor\" href=\"#动态计算图的优势与劣势\" aria-label=\"跳转到标题：动态计算图的优势与劣势\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>动态计算图的优势与劣势\u003C\u002Fh3>\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw03.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=EIxakjFg2tfQ2ugzuZKMvaBoJWE%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch3 id=\"pytorch-中的图优化\">\u003Ca class=\"anchor\" href=\"#pytorch-中的图优化\" aria-label=\"跳转到标题：PyTorch 中的图优化\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>PyTorch 中的图优化\u003C\u002Fh3>\u003Cul>\n\u003Cli>前端：在保持 Python 动态语义的前提下，对张量相关计算进行捕获与抽象，生成可优化的中间表示（IR）\u003C\u002Fli>\n\u003Cli>后端：基于中间表示进行图级优化、调度与代码生成\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch4 id=\"1-torchscript\">\u003Ca class=\"anchor\" href=\"#1-torchscript\" aria-label=\"跳转到标题：1. TorchScript\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>1. TorchScript\u003C\u002Fh4>\u003Cul>\n\u003Cli>介绍：通过\u003Ccode>torch.jit\u003C\u002Fcode>实现\u003C\u002Fli>\n\u003Cli>两个接口：\u003Cul>\n\u003Cli>\u003Ccode>torch.jit.trace\u003C\u002Fcode>\u003C\u002Fli>\n\u003Cli>\u003Ccode>torch.jit.script\u003C\u002Fcode>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw04.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=lkEGz8fjErj1EWPdsq37V0gBUFo%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch4 id=\"2-torch-fx\">\u003Ca class=\"anchor\" href=\"#2-torch-fx\" aria-label=\"跳转到标题：2. torch.fx\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>2. torch.fx\u003C\u002Fh4>\u003Cul>\n\u003Cli>介绍：Python 层 IR 与图变换工具\u003C\u002Fli>\n\u003Cli>技术亮点：\u003Cul>\n\u003Cli>\u003Cp>\u003Cstrong>Symbolic Tracing\u003C\u002Fstrong>：通过 Proxy 对象代替输入张量，记录所有张量运算的结构与路径。\u003C\u002Fp>\n\u003C\u002Fli>\n\u003Cli>\u003Cp>\u003Cstrong>fx IR\u003C\u002Fstrong>：高度简洁，6 种核心指令，\u003Ccode>placeholder\u003C\u002Fcode>、\u003Ccode>get_attr\u003C\u002Fcode>、\u003Ccode>call_function\u003C\u002Fcode>、\u003Ccode>call_module\u003C\u002Fcode>、\u003Ccode>call_method\u003C\u002Fcode>、\u003Ccode>output\u003C\u002Fcode>。\u003C\u002Fp>\n\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw05.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=IUHD5QHYKQTBhy3nk2WjM%2BGjjGA%3D\" alt=\"\">\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>Python 源到源转换\u003C\u002Fstrong>：\u003Cul>\n\u003Cli>IR 反向映射回 Python 源码\u003C\u002Fli>\n\u003Cli>基于 FX Graph 修改模型结构，导出新的\u003Ccode>nn.Module\u003C\u002Fcode>代码\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch4 id=\"3-pytorch-2-0-时代-torch-compile\">\u003Ca class=\"anchor\" href=\"#3-pytorch-2-0-时代-torch-compile\" aria-label=\"跳转到标题：3. PyTorch 2.0 时代：torch.compile\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>3. PyTorch 2.0 时代：torch.compile\u003C\u002Fh4>\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw06.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=gwK1wQapAQ8%2F7OYBBgBoSFt7Th8%3D\" alt=\"\">\u003C\u002Fp>\n\u003Cul>\n\u003Cli>介绍：统一编译接口，在尽量保留 Python 动态语义的前提下，通过图捕获与优化提升执行效率\u003C\u002Fli>\n\u003Cli>组件与步骤：\u003Cul>\n\u003Cli>\u003Cstrong>TorchDynamo（前端）\u003C\u002Fstrong>：解释器级 tracing 框架，在 Python 字节码层面捕获张量相关计算，生成受 guard 约束的 FX Graph\u003Cbr>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw07.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=3%2BNpWAxNPJQzFPHSl0TjyF3TZD0%3D\" alt=\"\">\u003Cul>\n\u003Cli>工作原理：拦截 Python 函数执行，记录张量相关 IR\u003C\u002Fli>\n\u003Cli>Guards 机制：监控变量类型与值，确保重用编译图\u003C\u002Fli>\n\u003Cli>Graph Break：子图划分，处理无法转换为静态图的部分\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw08.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=gp9DY2lub3I3VzRChN47wLt%2BTaM%3D\" alt=\"\">\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>AOT Autograd\u003C\u002Fstrong>：提前生成前向与反向联合计算图\u003Cul>\n\u003Cli>工作流程：Dispatcher 机制拦截与重定义 Tensor 操作，切分正反向计算图\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw09.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=7bOsjtz9tJ1WFto9F%2Bd%2FTlNRw1c%3D\" alt=\"\">\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>TorchInductor（后端）\u003C\u002Fstrong>：后端优化，生成高效可执行代码\u003Cul>\n\u003Cli>工作流程：接受 FX Graph，进行算子融合、内存规划与调度优化，并生成高效的执行代码\u003C\u002Fli>\n\u003Cli>示例：CUDA 设备上的 Triton 后端代码生成过程\u003C\u002Fli>\n\u003C\u002Ful>\n\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw10.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=mw0rbM9MpYqFP1QfgFJggnZJm3w%3D\" alt=\"\">\u003C\u002Fp>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_27_zw11.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=BZNZftc4LH5FHqAQxAvlyHqIT9g%3D\" alt=\"\">\u003C\u002Fp>\n\u003Ch3 id=\"总结\">\u003Ca class=\"anchor\" href=\"#总结\" aria-label=\"跳转到标题：总结\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>总结\u003C\u002Fh3>\u003Cp>本节课讲了 PyTorch 图优化技术体系的演进历程与核心特性。后续课程将讲解 AI 编译器中的后端优化。\u003C\u002Fp>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_01_zw6.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=DEth0OhERyx68xufLNiv%2BIt7foE%3D\" alt=\"2025冬季训练营课程表\">\u003C\u002Fp>\n\u003Ch3 id=\"学习贴士\">\u003Ca class=\"anchor\" href=\"#学习贴士\" aria-label=\"跳转到标题：💡 学习贴士\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>💡 学习贴士\u003C\u002Fh3>\u003Cp>更多细节及完整课程内容，请观看直播或查看课程回放。\u003C\u002Fp>\n\u003Cp>链接：\u003Ca href=\"https:\u002F\u002Fwww.infinitensor.com\u002Fcamp\u002Fwinter2025\">https:\u002F\u002Fwww.infinitensor.com\u002Fcamp\u002Fwinter2025\u003C\u002Fa>\u003C\u002Fp>\n\u003Ch3 id=\"训练营报名持续进行中\">\u003Ca class=\"anchor\" href=\"#训练营报名持续进行中\" aria-label=\"跳转到标题：📚 训练营报名持续进行中\">\u003Cspan class=\"octicon octicon-link\">\u003C\u002Fspan>\u003C\u002Fa>📚 训练营报名持续进行中\u003C\u002Fh3>\u003Cp>扫码报名与加入群聊，或点击 \u003Cstrong>「阅读原文」\u003C\u002Fstrong> 直达官网报名\u003C\u002Fp>\n\u003Cp>\u003Cimg src=\"https:\u002F\u002Foss.infinitensor.com\u002Fprod\u002Fimages\u002FKnowledge_Base\u002FKB_01_zw7.jpg?Expires=1792752557&OSSAccessKeyId=LTAI5t9R6uzvTa69CT92cFvV&Signature=OXZH4IYHRMHbFPG7EYCBYHhSMto%3D\" alt=\"\">\u003C\u002Fp>\n",[128,131,133,136,140,143,146,148,151],{"id":129,"text":130,"level":9},"pytorch-基础介绍","PyTorch 基础介绍",{"id":132,"text":132,"level":9},"动态计算图的优势与劣势",{"id":134,"text":135,"level":9},"pytorch-中的图优化","PyTorch 中的图优化",{"id":137,"text":138,"level":139},"1-torchscript","1. TorchScript",4,{"id":141,"text":142,"level":139},"2-torch-fx","2. torch.fx",{"id":144,"text":145,"level":139},"3-pytorch-2-0-时代-torch-compile","3. PyTorch 2.0 时代：torch.compile",{"id":147,"text":147,"level":9},"总结",{"id":149,"text":150,"level":9},"学习贴士","💡 学习贴士",{"id":152,"text":153,"level":9},"训练营报名持续进行中","📚 训练营报名持续进行中"]