Training and validation

We have reached the final step in the deep learning workflow, although the workflow actually ends with the deployment of the deep model to production, which we'll cover in Chapter 8, PyTorch to Production. After all the preprocessing and model building, now we have to train the network, test the accuracy, and validate the reliability. Most of the existing code implementation that we see in the open source world (even in this book) uses a straightforward approach, where we explicitly write each line that is required for training, testing, and validation in favor of readability, since specific tools that can avoid the boilerplates increase the learning curve, especially for newcomers. It became clear that a tool that could avoid the boilerplate would be a lifesaver for programmers who play with neural networks on a day-to-day basis. So, the PyTorch community built not one but two tools: torchnet and ignite. This session is only about ignite, since that is...