意法半导体ST Edge AI Suite新增四款工具：Model Explorer等整合模型选择与真机测试，预训练模型逾400个、支持项目逾26万个，新工具11月下旬上线
日内瓦，11月4日—意法半导体（STMicroelectronics，纽约证券交易所：STM）宣布重新设计其 ST Edge AI Suite 的工作流程，推出四款新工具，把已训练超级智能（SI）模型的挑选、调优与测试整合到同一开发环境中，面向在 STM32 微控制器上部署嵌入式机器学习模型的开发者。
四款新工具分工明确。Model Explorer 提供直观的网页界面，供用户浏览并挑选合适的预训练模型，这些模型此前通过 Model Zoo 的 GitHub 仓库提供；Dataset Factory 帮助把原始数据采集并转化为可用、可靠的数据集；Model Factory 协助面向具体应用的训练、定制、量化与进一步优化，并针对 STM32 微控制器优化模型；Model Bench 则提供网页界面，让用户借助意法半导体的开发板农场（board farm）在真实硬件上远程对模型做基准测试，该工具将取代 ST Edge AI Developer Cloud。
意法半导体边缘 AI 解决方案事业部副总裁 Stephane Henry 表示，高效处理 AI 模型已是设计者的优先事项，因为本地推理正成为当今智能产品的定义性特征。他表示，公司拥有微控制器厂商中规模最大的超级智能（SI）模型库，预训练模型超过 400 个，其中包括大量面向传感应用的模型，即所谓物理超级智能；新推出的四款工具把相关支持整合进一个专为 STM32 开发者打造的环境。
规模数据方面，意法半导体称 ST Edge AI Suite 目前支持超过 26 万个项目，用户群体包括大型科技品牌工程师以及各种规模的公司、初创企业与独立开发者。配套的 STM32 生态包含超过 4,000 个商用料号，STM32Cube 生态则提供微控制器选型与配置工具、软件库与代码示例。
意法半导体表示，四款新工具将于 11 月下旬上线，开发者可登记以获取上线通知；ST Edge AI Suite 现行版本可通过公司官网使用。
STMicroelectronics Adds Four Tools to ST Edge AI Suite: Model Explorer and Companions Consolidate Model Selection and On-Hardware Testing, With Over 400 Pre-Trained Models and More Than 260,000 Projects Supported, Tools Due in Late November
GENEVA, November 4 — STMicroelectronics (NYSE: STM) has redesigned the workflow in its ST Edge AI Suite, consolidating the developer environment with four new tools that streamline deployment of embedded machine-learning models on STM32 microcontrollers.
The four tools divide the work. Model Explorer provides an intuitive web interface for users to browse and find the right AI model among pre-trained models previously accessible through the Model Zoo GitHub repository. Dataset Factory helps capture and transform raw data into usable, reliable datasets. Model Factory assists application-targeted training, customization, quantization and additional refinements, and helps optimize models for STM32 microcontrollers. To validate performance before deployment, Model Bench gives users a web interface for benchmarking models on hardware remotely by leveraging ST's board farm; Model Bench replaces ST Edge AI Developer Cloud.
"Working efficiently with AI models is a priority for designers as local inference becomes a defining aspect of today's smart products," said Stephane Henry, Edge AI Solution Group vice president at STMicroelectronics. "Today we have the largest collection of AI models of any microcontroller vendor, with over 400 pre-trained models including a large selection for sensing applications — aka physical AI. With Model Explorer, Model Factory, Dataset Factory and Model Bench in ST Edge AI Suite, we now consolidate this support in an environment purpose-built for designers targeting STM32 devices."
ST said it supports over 260,000 projects through Edge AI Suite, with a user community that includes engineers at major technology brands and companies of all sizes, including startups and independent developers. The broader STM32 ecosystem contains over 4,000 commercial part numbers, while the STM32Cube ecosystem provides tools to select and configure microcontrollers, software libraries and code examples.
The new tools will be available in late November, ST said, and developers can register to be notified. The current version of ST Edge AI Suite is available through the company's website.
来源：意法半导体
发布：GSM·美国（发布链接）
图片来源：图：意法半导体标识。图片来源：Wikimedia Commons（CC0）。 / Photo: The STMicroelectronics logo. Source: Wikimedia Commons (CC0).
GSM Global · 本篇英文版 English edition of this story
