Training scripts

[cryptexis]

Hello @zRzRzRzRzRzRzR,

first of all thank you for the extraordinary work. I have couple of questions related to the project:

1. In the resources there is a readme file on how to how to start the training script for the control_net, but the actual script is missing.
2. Also I was wondering if you could prioritize releasing the training script in diffusers before other stuff?
   I have tried adapting SD3 training script to CogView pipeline - the first difference I have noticed is how VAE handles scale and shift parameters. However when I train a toy model my outputs get corrupted:

It seems there is a crucial information I am missing.

Hope to hear from you soon and thanks again for the great work!

[zRzRzRzRzRzRzR]

This script is still in PR, according to my testing, the code seems not completely correct, you can try the code in this PR.

huggingface/diffusers#10809

[cryptexis]

excellent, thank you @zRzRzRzRzRzRzR I am already getting much better results