Generate images and view attention maps
Experiment with and compare different tokenizers
Convert SD1.5 embedding to SDXL format