Summary

This chapter was all about how to build a basic pipeline for deep learning development. The system we have defined in this chapter is a very common/general approach followed by different sorts of companies, with slight changes. The benefit of starting with a generic workflow like this is that you can build a really complex workflow as your team/project grows on top of it.

Also, having a workflow in the early stage of development itself will make your sprints stable and predictable. Finally, the division between steps in the workflow helps with defining roles for the team members, setting deadlines for each step, trying to accommodate each of them in sprints efficiently, and executing the steps in parallel.

The PyTorch community is making different tools and utility packages to incorporate into the workflow. ignite, torchvision, torchtext, torchaudio, and so on are such examples. As the industry grows, we could see a lot of such tools emerging, which could be fitted into...