Genie: Show Me the Data for Quantization

This repository provides the PyTorch implementation of our CVPR 2023 paper "Genie: Show Me the Data for Quantization".

I, Emre Akbulut, added the data distillation visualization of GENIE by choosing a specific model, just after the distillation process. I also added the pipeline of GENIE to show the steps of the process. The original repository is here.

Getting Started

Installation

pip3 install -r requirements.txt

Usage

# Zero-Shot Quantization (PTQ)
python3 main.py --model_name=resnet18 --bit_w=4 --bit_a=4 --samples=1024 --val_path=IMAGENET_PATH/val

# or Few-Shot Quantization (PTQ)
python3 main.py --model_name=resnet18 --bit_w=4 --bit_a=4 --samples=1024 --train_path=IMAGENET_PATH/train --val_path=IMAGENET_PATH/val

Detailed descriptions of arguments are provided in main.py.

Citation

If you find this work useful for your research, please cite our paper:

@InProceedings{jeon2023genie,
    author    = {Jeon, Yongkweon and Lee, Chungman and Kim, Ho-young},
    title     = {Genie: Show Me the Data for Quantization},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    year      = {2023},
}

License

This project is released under the Apache 2.0 License.