SDXL

All images included in this dataset were voted as "Not solved" by the community in https://huggingface.co/spaces/OpenGenAI/open-parti-prompts. This means that according to the community the model did not generate an image that corresponds sufficiently enough to the prompt.

The following script was used to generate the images:

import torch
from datasets import Dataset, Features
from datasets import Image as ImageFeature
from datasets import Value, load_dataset

from diffusers import DDIMScheduler, DiffusionPipeline
import PIL


def main():
    print("Loading dataset...")
    parti_prompts = load_dataset("nateraw/parti-prompts", split="train")

    print("Loading pipeline...")
    ckpt_id = "stabilityai/stable-diffusion-xl-base-1.0"
    refiner_ckpt_id = "stabilityai/stable-diffusion-xl-refiner-1.0"

    pipe = DiffusionPipeline.from_pretrained(
        ckpt_id, torch_dtype=torch.float16, use_auth_token=True
    ).to("cuda")
    pipe.scheduler = DDIMScheduler.from_config(pipe.scheduler.config)
    pipe.set_progress_bar_config(disable=True)
    refiner = DiffusionPipeline.from_pretrained(
        refiner_ckpt_id,
        torch_dtype=torch.float16,
        use_auth_token=True
    ).to("cuda")
    refiner.scheduler = DDIMScheduler.from_config(refiner.scheduler.config)
    refiner.set_progress_bar_config(disable=True)

    seed = 0
    generator = torch.Generator("cuda").manual_seed(seed)

    print("Running inference...")
    main_dict = {}
    for i in range(len(parti_prompts)):
        sample = parti_prompts[i]
        prompt = sample["Prompt"]
        latent = pipe(
            prompt,
            generator=generator,
            num_inference_steps=100,
            guidance_scale=7.5,
            output_type="latent",
        ).images[0]
        image_refined = refiner(
            prompt=prompt,
            image=latent[None, :],
            generator=generator,
            num_inference_steps=100,
            guidance_scale=7.5,
        ).images[0]

        image = image_refined.resize((256, 256), resample=PIL.Image.Resampling.LANCZOS)
        img_path = f"sd_xl_{i}.png"
        image.save(img_path)
        main_dict.update(
            {
                prompt: {
                    "img_path": img_path,
                    "Category": sample["Category"],
                    "Challenge": sample["Challenge"],
                    "Note": sample["Note"],
                    "model_name": ckpt_id,
                    "seed": seed,
                }
            }
        )

    def generation_fn():
        for prompt in main_dict:
            prompt_entry = main_dict[prompt]
            yield {
                "Prompt": prompt,
                "Category": prompt_entry["Category"],
                "Challenge": prompt_entry["Challenge"],
                "Note": prompt_entry["Note"],
                "images": {"path": prompt_entry["img_path"]},
                "model_name": prompt_entry["model_name"],
                "seed": prompt_entry["seed"],
            }

    print("Preparing HF dataset...")
    ds = Dataset.from_generator(
        generation_fn,
        features=Features(
            Prompt=Value("string"),
            Category=Value("string"),
            Challenge=Value("string"),
            Note=Value("string"),
            images=ImageFeature(),
            model_name=Value("string"),
            seed=Value("int64"),
        ),
    )
    ds_id = "diffusers-parti-prompts/sdxl-1.0-refiner"
    ds.push_to_hub(ds_id)


if __name__ == "__main__":
    main()