Issues with GAN

[untold shuttle]

I've been working on a GAN that generated pokemon music for a school project for the past few months. I feel like I've gained a decent understanding in what I'm doing, but I have yet to yield any favorable results. I've tried googling things, yet nothing helps. I'm not sure if it's just a lack of training of the AI or if there's a logical mistake somewhere in there that needs to be changed. I would greatly appreciate if someone could take a moment or two to look through my code and see if there's a logical fallacy somewhere, or if I just need to run it for longer

[exotic pumice]

why not use diffusion?

[manic pond]

Main advantage of diffusion over GANs is that it is easy to use and easy to train. Bugs are usually not critical. You can probably learn a lot more in the course of figuring out how to training a GAN though..

[untold shuttle]

because i couldn't find anything that helped me learn how to make it myself, seemed like everything i found was premade and all you have to do it feed it some images

[exotic pumice]

u can try and make it urself too

tho ofc it is a bit more complicated

simply cuz of the fact that u have to add gaussian noise using some scheduler in a controlled way