Kitsuya0828 / DSFLplus

The implementation of DS-FL+ (Energy-based Knowledge Distillation for Communication-Efficient Federated Learning, IEICE 2024 student poster session).

DS-FL+

Note

The implementation of DS-FL+ (Energy-based Knowledge Distillation for Communication-Efficient Federated Learning, IEICE 2024 student poster session) is available here.

This is the implementation of DS-FL+ (Energy-based Thresholding and Knowledge Distillation for Communication-Efficient Federated Learning on Non-IID Data).

Requirements

• Python version: 3.10.13
• CUDA version: 11.6

Setup

git clone https://github.com/Kitsuya0828/DSFLplus.git
cd DSFLplus

conda env create -f env.yaml
conda activate dsflplus

# or

# https://pytorch.org/get-started/previous-versions/
conda create -n dsflplus python=3.10.13
conda activate dsflplus
pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 --extra-index-url https://download.pytorch.org/whl/cu116
pip install black fedlab gdown isort matplotlib mypy pandas plotly tensorboard

Run

cd src

python main.py --help

# or

python main.py \
--algorithm=dsflplus \
--task=cifar10 \
--partition=shards \
--num_shards_per_client=2 \
--seed=42 \
--public_private_split=even_class \
--private_size=40000 \
--public_size=10000 \
--public_size_per_round=1000 \
--sample_ratio=1.0 \
--com_round=1000 \
--temperature=0.1 \
--total_clients=100 \
--batch_size=100 \
--epochs=5 \
--lr=0.1 \
--kd_epochs=5 \
--kd_batch_size=100 \
--kd_lr=0.1 \
--ood_detection_score=energy \
--ood_detection_threshold_delta=0.0025 \
--test_batch_size=500 \
--comment=shards_energy_delta00025