Why are training and testing so slow

[InstantWindy]

Why are training and testing so slow?

[nuoran7607]

One possible reason is that most of the time is spent on reading images. Since there are so many images in ImageNet, it is best to use linux multiprocessing during training and testing (i.e. workers parameter is not set to 0, normal linux systems support workers parameter as 8). Another solution is to put training and test images on a solid-state drive, which can greatly increase the speed of reading images compared to a mechanical hard drive

[InstantWindy]

I did not read the ImageNet dataset, but read the Cifar dataset