effect is not good in other dataset

[huige555551]

changed the dataset of a drink and the dataset of the instant noodles. I felt that the effect was not very good. I used 1000epoch, d_oss: both converge to below 0.1, but g_loss: has been between 4-6, he At the beginning, it converges to below 1 but after 50epoch, it reaches 4-6. The hyperparameter is the same as your github project. Seek the guidance of the great God.

[lzzlxxlsz]

hello，after training the DCGAN ,the generated picture are kind of blurry，and they seem like the same color ,do you know the reason?