Make RepVGG Greater Again: A Quantization-aware Approach （AAAI2024）

📸 Release

⏳ QARepVGG training code. Note that the implementation is already provided in YOLOv6 and used in YOLO-NAS. Both are well known object detectors.

🦙 Model Zoo

Some models (continue updating)

Model	Checkpoint	Log
QARepVGG-B0	TBD	B0_log

🔔 Usage and License Notices: This project utilizes certain datasets and checkpoints that are subject to their respective original licenses. Users must comply with all terms and conditions of these original licenses. This project is licensed permissively under the MIT license and does not impose any additional constraints.

🛠️ Install

Clone this repository and navigate to QARepVGG folder

git clone  https://github.com/cxxgtxy/QARepVGG.git
cd QARepVGG

Install Package

conda create -n  QARepVGG python=3.10 -y
conda activate QARepVGG
pip install --upgrade pip
pip install -r requirements.txt

🗝️ Quick Start

QARepVGGBlockV2 is the default implementation, and we also provide other variants(for ablation and not recommended for use)
We use B0 for example, which is trained for 120 epochs on ImageNet 1k dataset.
```Shell
sh train_QAV2_B0.sh
```

🤝 Acknowledgments

RepVGG: the codebase we built upon. Thanks for their wonderful work! 👏
mmsegmentation: the great open-sourced framework for segmentation! 👏

✏️ Reference

If you find QARepVGG useful in your research or applications, please consider giving a star ⭐ and citing using the following BibTeX:

@inproceedings{chu2023make,
  title={Make RepVGG Greater Again: A Quantization-aware Approach},
  author={Chu, Xiangxiang and Li, Liang and Zhang, Bo},
  booktitle={AAAI},
  year={2024}
}

Sachitanmuo/QARepVGG_