baranyildirim/evogan

EvoGAN: Evolutionary Algorithm based Neural Architecture Search for Generative Adversarial Networks

• 👋 Hey, I’m Baran

evogan's Issues

Why isn't list comprehension used?

In the evo_train.py file there is a function called generate_new_dna where a for loop is used instead of list comprehension. I wanted to inform you that list comprehension is MUCH slower and less readable than for loops. See here.

I really like this project and want to use your code, but the fact that you used a for loop is honestly despicable. Please fix it and I will use your code and star it.

Questions about training details

Hello, I am very interested in your paper. I would like to learn more about the execution steps of the code. Can you describe the training process in detail?How to get the final structure searched out, and further train it to calculate its IS and FID values, thank you.

The corresponding IS and FID values could not be reached

Distinguished Researcher
Once I had configured the environment, I started evo_train.py after an afternoon of training. IS size IS 1.455, FID size IS None, what IS wrong with me?
I think there is something wrong with me. The network is not working. Looking forward to your reply, thank you very much

2022-03-19-18-55_train.log