Note: This video may require joining the NVIDIA Developer Program or login

GTC Silicon Valley-2019 ID:S9335:Metric Analysis and Performance Optimization in TensorFlow

Tong Yu(Inspur Corporaton)
We'll describe methods to improve the performance of FaceNet in parallel computing environments by quantitatively identifying major bottlenecks in performance and investigating the influence of measurements on accuracy and efficiency in TensorFlow. We'll discuss our performance monitoring tool for bottleneck analysis,and describe how data preprocessing, learning-rate scaling, and communication algorithms are incorporated into our neural network.

View the slides (pdf)