Deep convolution generative adversarial network

When should I stop training of DCGAN depend on train&val vs loss graph?