Too Long; Didn't Read
Using PyTorch, FastAI and the CIFAR-10 image dataset, we’ll try to replicate the approach used by the FastAI team to win the Stanford DAWNBench competition. We’re using multiple workers to leverage multi-core CPUs. The data is used to normalize the data and apply transformations faster. We use a model called WideResNet-ResNet, inspired from a family of networks introduced in the paper Resualual Networks. There are 50,000 training images (5,000 per class) and 10,000 test images.