PickScore v1 EU AI Act Compliance Profile

Model Card for PickScore v1

This model is a scoring function for images generated from text. It takes as input a prompt and a generated image and outputs a score. It can be used as a general scoring function, and for tasks such as human preference prediction, model evaluation, image ranking, and more. See our paper Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation for more details.

Model Details

Model Description

This model was finetuned from CLIP-H using the Pick-a-Pic dataset.

Model Sources [optional]

• Repository: See the PickScore repo
• Paper: Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation.
• Demo [optional]: Huggingface Spaces demo for PickScore

How to Get Started with the Model

Use the code below to get started with the model.

# import
from transformers import AutoProcessor, AutoModel

# load model
device = "cuda"
processor_name_or_path = "laion/CLIP-ViT-H-14-laion2B-s32B-b79K"
model_pretrained_name_or_path = "yuvalkirstain/PickScore_v1"

processor = AutoProcessor.from_pretrained(processor_name_or_path)
model = AutoModel.from_pretrained(model_pretrained_name_or_path).eval().to(device)

def calc_probs(prompt, images):
    
    # preprocess
    image_inputs = processor(
        images=images,
        padding=True,
        truncation=True,
        max_length=77,
        return_tensors="pt",
    ).to(device)
    
    text_inputs = processor(
        text=prompt,
        padding=True,
        truncation=True,
        max_length=77,
        return_tensors="pt",
    ).to(device)


    with torch.no_grad():
        # embed
        image_embs = model.get_image_features(**image_inputs

Your obligations as a PickScore v1 deployer

Track your progress

Sign up to track obligations, receive deadline reminders, and generate audit packages.

Tags

Library