Quantised Autoencoder with Attention around embedding

Table of config documentation

cl parsing arguments

TO DO:

• clear setup for training and sampling (variables that are not relevant)
• organise losses better
• add dropout
• add exponential moving average (ema)
  • https://github.com/CompVis/stable-diffusion/blob/21f890f9da3cfbeaba8e2ac3c425ee9e998d5229/ldm/modules/ema.py#L5
  • https://github.com/CompVis/stable-diffusion/blob/21f890f9da3cfbeaba8e2ac3c425ee9e998d5229/ldm/models/autoencoder.py#L55
  • https://github.com/CompVis/stable-diffusion/blob/21f890f9da3cfbeaba8e2ac3c425ee9e998d5229/ldm/models/diffusion/ddpm.py#L22