All the WGAN (and WGAN-GP) implementations that I have seen have a linear layer as the last layer of the critic. Also in my experience this seems to perform better than other activations (let’s say Tanh). Does anyone have any insights about why this is the case?