Please let me know - it seems impossible to get trained checkpoints for text to image.
I'll be sure to post an update! It's not a Diffusion Model, but I believe this VQGAN-CLIP implementation has weights available - https://github.com/nerdyrodent/VQGAN-CLIP