Download the dataset from here, rename the dictionary to dataset
and move it to the root of this project.
Example command to run local simulation of federated learning using FedAvg:
python server.py --simulated --rounds 50 --strategy fedavg --train_epochs 3
You can also run following command ./bin/run-proper-flower.sh
to run a real local federation with 3 clients using FedAvg with 50 rounds. Beware that this command is only working on macOS.
For additional documentation of all features of this example repo, feel free to dive into any of the root python files which all can be individually run. Also for simple docs you can add --help
flag for basic explanation of the scripts.
To browse the logs use command tensorboard --logdir=dir_xy
and replace dir_xy
accordingly:
classic-models/
- logs from non federated trainings will be logged therefederated-models/
- logs from federated trainings without using any simulated clientsfederated-models-simulated/
- logs from federated trainings with the use of simulated clients