No convergence in onw dataset

[ghost]

Hello
I'm trying to train WGAN in order to increase the amount of images (3211 img) that I have from my own data-set. The size of my images are 1242x375, but only for testing purposes of this net I am resizing them to 256x256.
To add my own data-set, I just added this code lines to the main.py file:

elif opt.dataset == 'own': normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225]) dataset = dset.ImageFolder(root=opt.dataroot, transform=transforms.Compose([ transformaciones.Transformaciones(), transforms.RandomHorizontalFlip(0.1), transforms.Resize((opt.imageSize,opt.imageSize)), transforms.ToTensor(), normalize, ]))

Transformaciones is the data augmentation module that i have build.
I launch the training with 6 extra layers and 800 epoch.
The thing is that generator net only produces noise. I tried to use the generator net when the results were like lossG: 0.08 approx, and I only had noise in the images.
I am doing something wrong? Is because I have only a few images to train?
Someone has tried to use an own data-set and managed to made it work?

Thanks!