When should I stop a training of WGAN model?

The loss function of the WGAN is a continuous one. It doesn't have a convergence point. I don't really understand when we should stop the training.