Training for Stage A?

[drscotthawley]

"We provide training scripts for both Stage B and Stage C"

I would actually like to train a Stage A.

Would that be performed using your VQGAN-pytorch repo, or the VQGAN training from the "Taming Transformers" paper, ....or is something else that would be better to use for training Stage A?

I've referred to Appendix E in your paper for hyperparameters and details, just also looking for code, as my current custom implementation may contain errors, so I want to compare with others'.

...I notice that your VQGAN-pytorch repo doesn't seem to contain any ConvNext layers, which goes against what I read in Appendix E of the Wuerstchen paper.

Thanks!