This one was version 1.0 further trained on labeled images from the same image dataset. It'll be easier to get a larger variety of images with this one compared to 1.0.

Note: you'll want to download the amazingarchitecture.zip file from the "Training Images" (which only contains the labels), unzip it to get the txt file that contains the labels to the training images, and extract it to the dynamic prompts folder for Automatic1111. Then add __amazingarchitecture__ to the prompt to get a random prompt which should increase the variety of images you can generate.