This is the official implementation for the CVPR2024 paper

pip install -r requirements.txt

• main.py: endpoint for starting experiments
• option.py: hyper-parameters for experiments
• trainer.py: includes three algorithms: "FedMPQ", "AQFL", "FP"
• QuantOptimizer.py: quantization-aware optimizer
• sampling.py: functions for generating data partitions with Dirichlet distribution
• ./client: includes clients implementing different algorithms
• ./server: includes server implementing aggregation
• ./model: includes ResNet model with bit-level operation
• ./utils: utility function for evaluation
• ./configs: includes training configurations for different experiments

python main.py --config ./configs/CIFAR10_FedMPQ_0.5.json 

This repository is built on the top of BSQ.

Please cite our paper, if you think this is useful:

@inproceedings{chen2024mixed,
  title={Mixed-precision quantization for federated learning on resource-constrained heterogeneous devices},
  author={Chen, Huancheng and Vikalo, Haris},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={6138--6148},
  year={2024}
}