How to use Spectral Normalization in RNN or Transformer?

Hi everyone, I am trying to train a GAN model with RNN and Transformer. Some paper show that using the spectral normalization is beneficial to training. However, The examples of torch.nn.utils.spectral_norm I saw are all used on CNN and Linear.

Can I use spectral normalization in RNN or Transformer? How to use?
Thanks.

Hello! I have the same question. Have your question been addressed?