Open
Description
Motivation and description
This has come up a couple of times now. Currently the relevant documentation isn't hidden, but it takes a few more steps to get there than it probably ought to. It also lacks some more unifying and advanced examples.
Possible Implementation
The baseline is to use what we currently have on the section on the layers page. Then we should layer on trainable
and a couple more complex examples. Finally, we should show how all this works for nested models through integration with the rest of Flux functionality (e.g. Optimisers). Somewhere in there should also be a hint to check ot the built-in layers first so people aren't inadvertently reinventing the wheel.