Hi there! In these three tutorials, we show you how to get started with
torch – whether you’ve done deep learning before (with Keras, for example) or not.
Guess the correlation is an end-to-end example that starts with data loading, proceeds by creating a neural network, and ends with model training and evaluation. It was written to be, at the same time, instructive to those who come with prior deep learning experience and accessible to an audience new to this topic. (OK - it is also meant to be a nice and interesting use case :-))
That said, if you’re part of the latter group, you’ll probably still want to get some more general background on neural networks. For some recommendations, you could take a look at the Resources page. And please don’t get hung up on details for now – just follow the example’s general flow to get a feeling for how it is to train a neural network.
Next, What if? Experiments and adaptations answers some more general questions you may have had following the first tutorial, and invites you to perform a few experiments. This page is mostly intended for newcomers, but may well be worth skimming over even for more experienced users (and be it just for the code snippets).
Finally, Create your own dataset will be of practical importance for everyone using
torch. That’s because unlike in our introductory example, you won’t always be able to work with data already packaged so it can directly be fed into a
DataLoader. Creating a custom
Dataset isn’t difficult – it’s just something you need to be able to do.
And once you’ve gone through all this? For one, it would be useful to get comfortable with some of the basic building blocks. Check out the series of tutorials in Get technical for in-depth information about tensors, automatic differentiation, modules, and more. In parallel, why not take a look at our gallery of applied examples and see the variety of tasks that can be approached with