Pytorch implementation of TRS

This is the implementation of the paper "Scheduling Weight Transitions for Quantization-Aware Training".

For detailed information, please checkout the project site [website] or the paper [arXiv].

Requirements

This repository has been tested with the following libraries:

• Python 3.8.8
• Numpy 1.19.2
• PyTorch 1.8.1
• Torchvision 0.9.1
• TensorBoard 2.5.0

How to run

We provide examples of commands in the run.sh file.
Training data (i.e., CIFAR-10/100) will be automatically downloaded.
You can modify the scripts for different settings (e.g., bit-widths).

Citation

@inproceedings{lee2024scheduling,
  title={Scheduling Weight Transitions for Quantization-Aware Training},
  author={Lee, Junghyup and Jeon, Jeimin and Kim, Dohyung and Ham, Bumsub},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  year={2025}
}