A short tutorial showing the input arguments for NLL loss/ cross entropy loss would be incredibly helpful

The arguments NLL loss (and by proxy cross entropy loss) take are in a relatively weird format. The documentation for the function does all it can within reason of the original documentation, but there's an incredible number of questions posted about weird problems giving them the kind of arguments. Much more than for other comparable things, and most don't really have good reusable answers.

The obvious solution to this is for someone to create a simple example based tutorial of using NLL loss, and clearly showing exactly what format the arguments need to be in (perhaps starting with input and targets that are one hot encoded to make it as idiot proof as possible).

I've spent 4 hours trying to solve a problem exactly like this without success, and am about to refer to source over it. Someone please take mercy on future programmers.

Review the existing NLL loss and cross entropy documentation and the repository's tutorial structure first. Create a small tutorial that demonstrates the required input and target formats, including the requested one-hot-encoded starting example, and consider the work done when common argument-shape questions are answered clearly.