I get this error when executing stable diffusion:

Hey I have am training VQGAN on custom dataset and wanted to know if there's is a way to look at the model performance while training apart from monitoring the loss, like generating images after each epoch and also where will the trained model be stored ?

how to train a images whose h and w different? All batch size values provided in conf file are small, seems set for one GPU?

how to train a images whose h and w different? not setting for h and w in the conf. are changes of the codes needed to achieve this?

set resolution and size to 1024 in conf file if I want to generate high resolution images? I encouter the problem of cuda out of memory, anybody achieve this?

any first stage model of celeba is available? links?

I conduct training transformer on custom dataset, it seems like the program will log images on the local disk, but I only got the images of the first five epochs. The training is still going, but it no longer log images, how to fix this problem?

I'd like to set the condion stage to be the same with the first stage, how to configure this in .yaml file?

Why pad input manually?