Training two models at once with differing schedules

[TheSeparatrix]

Hi again!

I am currently trying out a sort of adversarial autoencoder where a classifier is being trained on the auto encoder latent space while the autoencoder also trains regularly.
I am wondering if there is a way in PyTorch-Lightning to run some epochs of classifier training before each epoch of the autoencoder training.
Something like:
For each training epoch:
Train classifier for 20 epochs on current latent space.
THEN train autoencoder 1 epoch.

Can I somehow specify this elegantly in the model (maybe in training_step or in Trainer)?