SpaFL - NeruIPS 2024

This is a repository for the paper "SpaFL: Communication-Efficient Federated Learning with Sparse Models and Low Computational Overhead?" published in NeurIPS 2024

Install

1. create a conda env with python=3.10.14
2. pip install -r requirements.txt
3. pip install --upgrade protobuf wandb

Run

1. FMNIST: python train.py --learning_rate 0.001 --th_coeff 0.002 --local_epoch 5 --seed 1
2. CIFAR-10: python train.py --affix grad_clip_3_seed_1_conv4 --model conv4 --learning_rate 0.01 --th_coeff 0.00015 --batch_size 16 --alpha 0.1 --local_ep 5 --clip 3 --seed 1
3. CIFAR-10 with ViT: python train.py --model vit --batch_size 16 --comm_rounds 500 --learning_rate 0.01 --alpha 0.1 --th_coeff 0.0001 --local_epoch 1 --clip 3 --seed 1
4. CIFAR-100: python train.py --model resnet18 --comm_rounds 1500 --learning_rate 0.01 --lr_decay 0.993 --alpha 0.1 --th_coeff 0.0007 --local_epoch 7 --clip 15 --seed 1