24 May 2021 3 5K Report

Hi all,

I'm currently working on a project about hardware benchmark for deep learning capability. My Github for this project is as follows:

https://github.com/vohuynhquangnguyen/Deep-Learning-Hardware-Benchmark

My questions so far are:

  • What are good metrics to measure the hardware capability? Currently, I'm using the following metrics: total execution time, flops, and presence of errors. Are there any other relevant metrics that can be used as well? If there are, how do I implement them?
  • In this project, I implemented a couple of deep learning models for benchmarking purposes. How can I optimize my included these deep learning models more?
  • Similar questions and discussions