Vit L 14 CLIPA 336 Datacomp1b

🚀 ViT-L-14-CLIPA-336-datacomp1B Model Card

A CLIPA-v2 model for zero-shot image classification.

🚀 Quick Start

The ViT-L-14-CLIPA-336-datacomp1B is a CLIPA-v2 model designed for zero-shot image classification. It is trained on the mlfoundations/datacomp_1b dataset and can achieve high accuracy in zero-shot ImageNet classification.

✨ Features

• Model Type: Contrastive Image-Text, Zero-Shot Image Classification.
• Original Repository: https://github.com/UCSC-VLAA/CLIPA
• Dataset: mlfoundations/datacomp_1b
• Papers:
  • CLIPA-v2: Scaling CLIP Training with 81.1% Zero-shot ImageNet Accuracy within a $10,000 Budget; An Extra $4,000 Unlocks 81.8% Accuracy
  • An Inverse Scaling Law for CLIP Training

📦 Installation

Not provided in the original document, so this section is skipped.

💻 Usage Examples

Basic Usage

import torch
import torch.nn.functional as F
from urllib.request import urlopen
from PIL import Image
from open_clip import create_model_from_pretrained, get_tokenizer

model, preprocess = create_model_from_pretrained('hf-hub:ViT-L-14-CLIPA-336')
tokenizer = get_tokenizer('hf-hub:ViT-L-14-CLIPA-336')

image = Image.open(urlopen(
    'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
))
image = preprocess(image).unsqueeze(0)

text = tokenizer(["a diagram", "a dog", "a cat", "a beignet"], context_length=model.context_length)

with torch.no_grad(), torch.cuda.amp.autocast():
    image_features = model.encode_image(image)
    text_features = model.encode_text(text)
    image_features = F.normalize(image_features, dim=-1)
    text_features = F.normalize(text_features, dim=-1)

    text_probs = (100.0 * image_features @ text_features.T).softmax(dim=-1)

print("Label probs:", text_probs)  # prints: [[0., 0., 0., 1.0]]

📚 Documentation

The model details are as follows:

🔧 Technical Details

Not provided in the original document, so this section is skipped.

📄 License

The model is licensed under the apache-2.0 license.

📚 Citation

@article{li2023clipav2,
      title={CLIPA-v2: Scaling CLIP Training with 81.1% Zero-shot ImageNet Accuracy within a $10,000 Budget; An Extra $4,000 Unlocks 81.8% Accuracy}, 
      author={Xianhang Li and Zeyu Wang and Cihang Xie},
      journal={arXiv preprint arXiv:2306.15658},
      year={2023},
}

@inproceedings{li2023clipa,
      title={An Inverse Scaling Law for CLIP Training}, 
      author={Xianhang Li and Zeyu Wang and Cihang Xie},
      booktitle={NeurIPS},
      year={2023},
}