Tutorial about adaptative weights for adversarial training

[Warvito]

In the original code from taming transformers and latent diffusion models, the weight for the discriminator adversarial loss is defined by the ratio between the nll_loss and the g_loss gradients (https://github.com/CompVis/taming-transformers/blob/3ba01b241669f5ade541ce990f7650a3b8f65318/taming/modules/losses/vqperceptual.py#L63).

In our experiments and the paper on 3D brain images (https://arxiv.org/pdf/2209.07162.pdf), we were not able to make it work well on both, 2D and 3D data. However, this might be useful for others.

[danieltudosiu]

We have the code for it in the KCL's VQ-VAE codebase