Diffusion

For All PartiPrompts, its 1024 prompts, comparing V2-V1 and V1-V2 using llava (verbose captioning model) and blip (concise captioning model).

I forgot to visualize the images but you can see sample captions in the 'results' table