LightningFL

A research-oriented federated learning framework implemented with pytorch and WandB.

Follow the following steps, easy to reproduce the experiment results.

Step 0:

Register a WandB account. Refer to WandB QuickStart.

Step 1:

cd LightningFL/

pip install -r requirements.txt

Step 2:

cd LightningFL/algorithm/

python main.py --algorithm=fedmc --alpha=0.1 --batchSize=50 --clientsPerRound=10 --cuda=True --dataset=cifar10 --decayStep=1 --diffCo=1 --drop=big --epoch=5 --evalInterval=1 --lr=0.1 --lrDecay=0.99 --mode=concat --model=cifar10 --mu=0.0001 --numRounds=100 --omega=100 --seed=12

Step 3:

The experiments (accuracy and loss curves) can be seen in https://wandb.ai/.