📚 Documentation

Hi, I've come across this issue where it's explained how a LightningModule should be at the top level and nesting them isn't supported. #20053 (comment)

However, in the transfer learning docs, I can see this nesting being done.
https://github.com/Lightning-AI/pytorch-lightning/blob/1f5add327fd88fe288a2f889d720e5d5e06bd7d2/docs/source-pytorch/advanced/transfer_learning.rst

Is this nesting well defined now? How is CIFAR10Classifier supposed to behave if AutoEncoder uses self.log() or has a hook like predict_step() defined?

Thanks!

cc @lantiga @Borda