Summary

In this chapter, we first learned about PyTorch Mobile and how it can be used to convert traced PyTorch model artifacts into optimized model objects that can run on mobile devices. We then learned how to build an Android app that uses PyTorch Mobile to classify images of handwritten digit captured by a phone camera using a pre-trained MNIST model. We then repeated this exercise for iOS, where we built an iOS app, again from scratch, to classify images of handwritten digits into one of 10 classes. In the next chapter, we will discuss various tools and libraries such as fastai and PyTorch Lightning that speed up and simplify the process of model training in PyTorch. We will also learn how to profile PyTorch code to understand resource utilization, using PyTorch profiler.