This is a PyTorch wrapper for https://github.com/xiexi51/RTopK. The original code by Xie et al. has an MIT license. We cite it as follows:
@inproceedings{
xie2025rtopk,
title={RTop-K: Ultra-Fast Row-Wise Top-K Selection for Neural Network Acceleration on GPUs},
author={Xi Xie and Yuebo Luo and Hongwu Peng and Caiwen Ding},
booktitle={The Thirteenth International Conference on Learning Representations},
year={2025},
url={https://openreview.net/forum?id=PHg4rAXFVH}
}