DCGANs | Implementation of some basic GAN architectures in Keras | Machine Learning library

I am analyzing a meta-learning class that uses DCGAN + Reptile within the image generation.

I have two questions about this code.

First question: why during DCGAN training (line 74)

Following example is from here. This is an example training a GANs.

I'm trying to run code for a Deep Convolutional GAN from the official PyTorch site (https://pytorch.org/tutorials/beginner/dcgan_faces_tutorial.html) on my Mac.

When I try loading the data, I keep getting a "FileNotFound" error.

I am trying to use this version of the DCGAN code (implemented in Tensorflow) with some of my data. I run into the problem of the discriminator becoming too strong way too quickly for generator to learn anything.

Now there are some tricks typically recommended for that problem with GANs:

• batch normalisation (already there in DCGANs code)

• giving a head start to generator.

I did some version of the latter by allowing 10 iterations of generator per 1 of discriminator (not just in the beginning, but throughout the entire training), and that's how it looks:

Adding more generator iterations in this case helps only by slowing down the inevitable - discriminator growing too strong and suppressing the generator learning.

Hence I would like to ask for an advice on whether there is another way that could help the problem of a too strong discriminator?

I want to use condition GANs with the purpose of generated images for one domain (noted as domain A) and by having input images from a second domain (noted as domain B) and the class information as well. Both domains are linked with the same label information (every image of domain A is linked to an image to domain B and a specific label). My generator so far in Keras is the following:

I am working on GANs and decided to implement my algorithm using HyperGAN. Its a wrapper on DCGANs using TensorFlow. HyperGAN saves the output using TF's checkpoint method.

Later, I tried to run the load the model using:

I am trying to adjust this DCGAN code to be able to work with 2x80 data samples.

All generator layers are tf.nn.deconv2d other than h0, which is ReLu. Generator filter sizes per level are currently:

Running the DCGANS-tensorflow tutorial, more precisely lines python download.py mnist celebA and python main.py --dataset celebA --input_height=108 --train --crop, I get the following error: