How to freeze generator during training?

[the-entire-country-of-ireland]

How do I freeze all the generator weights during training, while still training the discriminator?

I tried setting do_Dmain, do_Gmain, do_Gpl, do_Dr1 = True, False, False, True within accumulate_gradients() of loss.py, but this caused the generator to diverge.
https://github.com/NVlabs/stylegan2-ada-pytorch/blob/main/training/loss.py#L59

Apart from accumulate_gradients() within loss.py, what else do I need to change to prevent the generator from continuing to train?

These are the faces produced after resuming training from FFHQ for 40 epochs with the above flags set.