Code for paper: [https://arxiv.org/abs/2301.09269]
Need manually change quantization type and network architecture
To use count sketch, please install https://github.com/nikitaivkin/csh
The parameter details of the model we used are shown in model_details.png