How to launch a separate fit within fit?

[RylanSchaeffer]

I want to train a network, and then after every K (a hyperparameter) epochs, attach a linear layer to the network and finetune the linear layer on a separate task (i.e. train and test using different dataloaders) for some number of epochs before returning to the main training loop. Is there any easy way to do this within PyTorch Lightning? The key challenge is doing this every K epochs, not before/after all epochs or before/after every epoch.