ARM 芯片的 Python + AI 算力优化 主讲人： 朱宏林 – 阿里云程序语言与编译器团队 简介 • 当今开发者们大量使用 Python 语言编写的 AI 程序。过去这些程序总跑在 GPU 或者 x86 架构的 CPU 上。然而综合考虑到功耗、成本、性能等因素，云厂商们开始建设 ARM 架构的服务平台，如 何整合 Python + AI 的相关软件并使其在该平台上发挥最高的性能成为了工程师们关注的焦点。 提升明显。目前，该成果已经被集成进 OpenBLAS 和 PyTorch 中。 • 本次演讲，将向大家介绍我们在倚天 710 ARM 芯片上开展的 Python + AI 优化工作，以及在 ARM 云平台上部署 Python + AI 任务的最佳实践。 深度学习 • 广泛使用的深度学习框架 • TensorFlow、PyTorch • 结合硬件（ARM 服务端芯片） • 倚天 710 •

Code 远程开发 —— 远程机器 VS Code 远程开发 —— WSL 垃圾分类 垃圾分类的不同目标用户 • ToC: 微信小程序 • ToB: AI on the Edge ToC: 微信小程序 三种 Azure 的 AI 解决方案 Azure Machine Learning • Your data • Your model Azure Custom Vision Vision • Your data • Our model Azure Computer Vision • Our data • Our model 解决方案 ToB: AI + IoT https://github.com/jxtxzzw/Microsoft-Hackathon-2019-Intelligent-Garbage-Classification-IoT-Edge-Solution • 数据隐私 • 成本 • 大规模部署 • 离线运行 • 支持不同的设备 • Linux, Windows • AMD64, ARM32v7, ARM64 Azure IoT Edge + AI 把 AI 运算下放到边缘计算节点。 如何运作？ 问题迎刃而解 • 延迟 • 数据隐私 • 成本 • 大规模部署 • 离线运行 • 支持不同的设备 • Linux, Windows • AMD64

26 Ø 目前赛灵思的AI开发工具支持 python 语言 输入，可以做 AI inference 的数据前处理 和后处理的工作。这部分工作可以在SOC 的ARM 里面实现。 Ø Xilinx AI Developer Resource Ø https://www.xilinx.com/products/ design-tools/ai-inference/ai- developer-hub Xilinx Edge AI Tutorial Ø https://github.com/Xilinx/Edge-AI- Platform-Tutorials Ø Xilinx ModelZoo Ø https://www.xilinx.com/bin/public /openDownload?filename=all_mo dels_20190528.zip Ø Xilinx AI Forums Forums Ø https://forums.xilinx.com/t5/Deep hi-DNNDK/bd-p/Deephi Python in Edge AI Computing – Xilinx DNNDK 27 马达控制和可预测性维护 FOC Algorithm HLS/SDSoC Translation and Overlay 马达控制可视化： Python with Jupyter

26 Ø 目前赛灵思的AI开发工具支持 python 语言 输入，可以做 AI inference 的数据前处理和 后处理的工作。这部分工作可以在SOC的 ARM 里面实现。 Ø Xilinx AI Developer Resource Ø https://www.xilinx.com/products/desig n-tools/ai-inference/ai-developer- hub Xilinx Edge AI Tutorial Ø https://github.com/Xilinx/Edge-AI- Platform-Tutorials Ø Xilinx ModelZoo Ø https://www.xilinx.com/bin/public/ope nDownload?filename=all_models_2019 0528.zip Ø Xilinx AI Forums Forums Ø https://forums.xilinx.com/t5/Deephi- DNNDK/bd-p/Deephi Python in Edge AI Computing – Xilinx DNNDK 27 马达控制和可预测性维护 FOC Algorithm HLS/SDSoC Translation and Overlay 马达控制可视化： Python with Jupyter

26 ➢ 目前赛灵思的AI开发工具支持 python 语言 输入，可以做 AI inference 的数据前处理和 后处理的工作。这部分工作可以在SOC的 ARM 里面实现。 ➢ Xilinx AI Developer Resource ➢ https://www.xilinx.com/products/desig n-tools/ai-inference/ai-developer- hub Xilinx Edge AI Tutorial ➢ https://github.com/Xilinx/Edge-AI- Platform-Tutorials ➢ Xilinx ModelZoo ➢ https://www.xilinx.com/bin/public/ope nDownload?filename=all_models_2019 0528.zip ➢ Xilinx AI Forums Forums ➢ https://forums.xilinx.com/t5/Deephi- DNNDK/bd-p/Deephi Python in Edge AI Computing – Xilinx DNNDK 27 马达控制和可预测性维护 FOC Algorithm HLS/SDSoC Translation and Overlay 马达控制可视化： Python with Jupyter

it is recommended only for development use). Tornado will also run on Windows, although this configuration is not officially supported and is recommended only for development use. Without reworking Tornado between Twisted and Tornado Utilities tornado.autoreload — Automatically detect code changes in development tornado.log — Logging support tornado.options — Command-line parsing tornado.stack_context — Exception reloading If you pass debug=True to the Application constructor, the app will be run in debug/development mode. In this mode, several features intended for convenience while developing will be enabled

programming. While some support of WSGI is available in the tornado.wsgi module, it is not a focus of development and most applications should be written to use Tornado’s own interfaces (such as tornado.web) directly it is recommended only for development use). Tornado will also run on Windows, although this configuration is not officially supported and is recommended only for development use. Without reworking Tornado reloading If you pass debug=True to the Application constructor, the app will be run in debug/development mode. In this mode, several features intended for convenience while developing will be enabled

programming. While some support of WSGI is available in the tornado.wsgi module, it is not a focus of development and most applications should be written to use Tornado’s own interfaces (such as tornado.web) directly it is recommended only for development use). Tornado will also run on Windows, although this configuration is not officially supported and is recommended only for development use. Without reworking Tornado reloading If you pass debug=True to the Application constructor, the app will be run in debug/development mode. In this mode, several features intended for convenience while developing will be enabled

it is recommended only for development use). Tornado will also run on Windows, although this configuration is not officially supported and is recommended only for development use. Without reworking Tornado reloading If you pass debug=True to the Application constructor, the app will be run in debug/development mode. In this mode, several features intended for convenience while developing will be enabled reload itself when anything changes. This reduces the need to manually restart the server during development. However, certain failures (such as syntax errors at import time) can still take the server down

programming. While some support of WSGI is available in the tornado.wsgi module, it is not a focus of development and most applications should be written to use Tornado’s own interfaces (such as tornado.web) directly it is recommended only for development use). Tornado will also run on Windows, although this configuration is not officially supported and is recommended only for development use. Without reworking Tornado between asyncio and Tornado Utilities tornado.autoreload — Automatically detect code changes in development tornado.concurrent — Work with Future objects tornado.log — Logging support tornado.options — Command-line