[{"data":1,"prerenderedAt":163},["ShallowReactive",2],{"content-query-6ZYQ6KczJS":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"body":13,"_type":157,"_id":158,"_source":159,"_file":160,"_stem":161,"_extension":162},"/news/zh/430","zh",false,"","TinyMS新工具发布！","TinyMS是一款主要用PyThon语言编写的开源深度学习开发工具包，基于以MindSpore为代表的新型开源深度学习框架，提供面向从数据准备到模型部署全流程的极简易用的高阶API封装，并通过易于扩展的模块化设计，提供覆盖多种业务场景的能力。","2021-03-31","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2021/03/31/4203bcab55934b88a51da7d451ca2bfb.png","news",{"type":14,"children":15,"toc":154},"root",[16,24,34,43,47,52,57,62,70,75,80,85,93,98,103,108,116,121,126,131,136,141,149],{"type":17,"tag":18,"props":19,"children":21},"element","h1",{"id":20},"tinyms新工具发布",[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":23,"value":33},"TinyMS项目简介",{"type":17,"tag":25,"props":35,"children":36},{},[37],{"type":17,"tag":38,"props":39,"children":42},"img",{"alt":40,"src":41},"image.png","https://bbs-img-cbc-cn.obs.cn-north-1.myhuaweicloud.com/data/forums/attachment/forum/202103/29/144845bxnvz9naqewxqr80.png",[],{"type":17,"tag":25,"props":44,"children":45},{},[46],{"type":23,"value":9},{"type":17,"tag":25,"props":48,"children":49},{},[50],{"type":23,"value":51},"TinyMS主要由data, model, serving等模块组成，分场景分领域提供transform数据预处理算子，复用MindSpore原生数据集提供常用数据集，如：cifar-10等。data提供部分自定义数据集和常用的数据下载和解压等常用工具集，model提供常用的预置模型，并提供模型构建，模型编译，模型训练、验证与推理。serving通过搭建服务器来提供AI模型应用服务,为新手提供快速推理的体验。",{"type":17,"tag":25,"props":53,"children":54},{},[55],{"type":23,"value":56},"TinyMS面向的主要用户群体为深度学习初学者、研究领域涉及深度学习结合的科研人员、以及深度学习相关业务应用开发的企业开发人员。",{"type":17,"tag":25,"props":58,"children":59},{},[60],{"type":23,"value":61},"通过搭配完整的在线课程教学，TinyMS提供目前业界最佳的深度学习入门与开发体验。",{"type":17,"tag":25,"props":63,"children":64},{},[65],{"type":17,"tag":29,"props":66,"children":67},{},[68],{"type":23,"value":69},"TinyMS vs Keras",{"type":17,"tag":25,"props":71,"children":72},{},[73],{"type":23,"value":74},"keras 是一个用 Python 编写的高级神经网络 API，将把用户体验放在首要位置，支持短时间内出实验。",{"type":17,"tag":25,"props":76,"children":77},{},[78],{"type":23,"value":79},"keras项目可以说是“大而全”，主要由dataset, layer, model和backend模块构成，提供较多常用的预置数据集，并分场景分领域提供数据预处理函数，layer网络层提供较完善，如：convolution卷积层，embedding嵌入层，pooling池化层等。backend支持多个后端（TensorFlow,CNTK和Theano），与TensorFlow版本不强耦合。Model提供模型选择（sequential）、网络层构建（输入层、输出层和池化层等），模型编译，模型训练、验证与推理。",{"type":17,"tag":25,"props":81,"children":82},{},[83],{"type":23,"value":84},"TinyMS在高阶API方面会更为简单抽象，较keras来说复杂度更低，比如提供了只需一行代码即可完成数据集的预处理，而且在设计中重点考虑到了Keras尚未提供单独好用的工具库，以及尚未提供的快速部署推理模块等。",{"type":17,"tag":25,"props":86,"children":87},{},[88],{"type":17,"tag":29,"props":89,"children":90},{},[91],{"type":23,"value":92},"TinyMS vs Fastai",{"type":17,"tag":25,"props":94,"children":95},{},[96],{"type":23,"value":97},"Fastai是为了帮助新手快速轻松出结果的高阶API项目, 其基于PyTorch的深度学习库，利用底层PyTorch库的灵活性，分领域分场景地提供包括对vision，text，tabular和collab模型的“开箱即用”的支持，后端对PyTorch的版本要求紧耦合。",{"type":17,"tag":25,"props":99,"children":100},{},[101],{"type":23,"value":102},"fastai 深度学习库项目较轻便，目录清晰易理解，可以说是“小而美”，主要由data, models和learner三大模块构成，其中，data提供了transform类方便开发者进行数据预处理操作。models按应用领域，提供部分预置网络，如：unet，快速实现模型构建。learner实现数据和模型的关联，并定义了一系列回调函数，帮助开发者快速厘清深度学习的架构，提供模型训练、模型评估、模型保存与加载和模型推理。除此之外，还拥有较丰富好用的工具集，如：数据下载，解压，图片验证和文件处理等工具函数。",{"type":17,"tag":25,"props":104,"children":105},{},[106],{"type":23,"value":107},"TinyMS在高阶API方面理念与Fastai相近，不同点在于TinyMS提供了常用的MindSpore预置数据集，方便开发者简化对数据集的调用，而且提供了Fastai尚未提供的快速部署推理模块等。",{"type":17,"tag":25,"props":109,"children":110},{},[111],{"type":17,"tag":29,"props":112,"children":113},{},[114],{"type":23,"value":115},"TinyMS开源社区简介",{"type":17,"tag":25,"props":117,"children":118},{},[119],{"type":23,"value":120},"TinyMS开源社区中除了TinyMS项目外，还有如下一些项目和活动：",{"type":17,"tag":25,"props":122,"children":123},{},[124],{"type":23,"value":125},"l Specification项目：主要用来协作制定面向模型训练脚本的格式规范。由于TinyMS提供了较为高阶的API抽象，因此诞生了ModelZoo脚本规范性和标准化的需求，便于高阶封装的持续迭代",{"type":17,"tag":25,"props":127,"children":128},{},[129],{"type":23,"value":130},"l tinyms-ai.github.io：开源实现的简单官方网站搭建，基于Github Page",{"type":17,"tag":25,"props":132,"children":133},{},[134],{"type":23,"value":135},"l RustedAI Team：目前只有组织成员可见，RustedAI是TinyMS旨在推动利用Rust语言编写更多的低运行时开销的深度学习组件。",{"type":17,"tag":25,"props":137,"children":138},{},[139],{"type":23,"value":140},"l 社区活动：我们会不定期的组织TinyMS模型拉力赛，以及多种多样的Meetup活动",{"type":17,"tag":25,"props":142,"children":143},{},[144],{"type":17,"tag":29,"props":145,"children":146},{},[147],{"type":23,"value":148},"TinyMS与开发者",{"type":17,"tag":25,"props":150,"children":151},{},[152],{"type":23,"value":153},"TinyMS是一个新生的开源项目，我们站在Keras、fastai等巨人的肩膀上，虽然在设计理念上有所创新，但依然需要社区开发者一起持续的协作，才能达到可以更好的服务学术界、产业界和开发者的深度和广度。",{"title":7,"searchDepth":155,"depth":155,"links":156},4,[],"markdown","content:news:zh:430.md","content","news/zh/430.md","news/zh/430","md",1776506092213]