摘要
为了客观、规范的评测一款深度学习训练框架的性能与功能,提出了一种较为科学的评测方法,在一定的约束条件下,对特定应用场景的关键指标进行检测,从而得出量化、客观的测试结果。并从深度学习训练框架的生态指标、易用性指标、性能指标及可靠性指标四个方面,进行了深度学习训练框架的评测方法研究。
In order to evaluate the performance and functions of a deep learning training framework objectively and normatively.Under certain constraints and clescribe a kind of scientific evaluation method to test the key indicators of specific application scenarios,and get the quantitative and objective test results.The evaluation method of deep learning training framework is studied from four aspects:ecological,usability,performance and reliability.
出处
《信息技术与标准化》
2020年第1期66-68,共3页
Information Technology & Standardization
关键词
人工智能
深度学习
训练框架
评测方法
artificial intelligence
deep learning
training framework
evaluation method