首页> 外文会议>Services conference federation;International conference on edge computing >Godec: An Open-Source Data Processing Framework for Deploying ML Data Flows in Edge-Computing Environments
【24h】

Godec: An Open-Source Data Processing Framework for Deploying ML Data Flows in Edge-Computing Environments

机译:GODEC:用于在边缘计算环境中部署ML数据流的开源数据处理框架

获取原文

摘要

We present Godec, a C++-based framework that allows easy transition of complex machine learning (ML) data flows to edge-computing environments where common data processing frameworks do not apply. Godec allows for free mixing of technologies such as Kaldi, TensorFlow and custom modules, all wrapped into a single OS process, making it easy to deploy inference engines on constrained environments like Android, iOS or embedded Linux. Godec achieves this by connecting the components into an arbitrary graph specified by a simple JSON file during startup. Despite being multithreaded, results between runs are guaranteed identical, allowing for immediate transition from offline experiments to deployment.
机译:我们提出了一个基于C ++的框架,可以轻松转换复杂的机器学习(ML)数据流到边缘计算环境,其中常见的数据处理框架不适用。 Godec允许自由混合kaldi,tensorflow和自定义模块,所有这些技术都包裹在一个OS过程中,使得在Android,iOS或嵌入式Linux等受限环境中易于部署推理引擎。 Godec通过将组件连接到启动期间由简单的JSON文件指定的任意图形来实现这一目标。尽管是多线程,但运行之间的结果保证相同,允许从离线实验转换到部署。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号