Details of the reproduction of the experimental results

[ZTzxj]

Hello, I would like to ask what is the prompt file for the validation set of the coco dataset used in your assessment, what validation set title did you use, and also how many images did you generate based on each of the prompts.

[horseee]

Hi, you can refer to the code snippet here:

def main(args):
    if args.dataset == 'parti':
        prompts = load_dataset("nateraw/parti-prompts", split="train")
    elif args.dataset == 'coco2017':
        dataset = load_dataset("phiyodr/coco2017")
        prompts = [{"Prompt": sample['captions'][0]} for sample in dataset['validation']]
    else:
        raise NotImplementedError