Performance benchmarking of CPU versus GPU models for training and inference

Now that we've done all that work, let's explore some of the advantages of using a GPU for deep learning. First, let's go through how to actually get your application to use CUDA, and then we'll go through some of the CPU and GPU speeds.

How to use CUDA

If you've completed all the previous steps to get CUDA working, then using CUDA is a fairly simple affair. You simply need to compile your application with the following:

go build -tags='cuda'

This builds your executable with CUDA support and uses CUDA, rather than the CPU, to run your deep learning model.

To illustrate, let's use an example we're already...