saadiabadi / cifer_example

this repository is for freezing layers experiment using cifer

FEDn implementation of cifar10 with keras

Setting up a client

Initiate your federation

This command loads and partitions the dataset, generates a docker-compose.yaml for your choice of numbers of clients and initiates a seed model to start from: (Replace: {NR_OF_CLIENTS} with the number of clients you want to build you federation with)

pip install -r init_requirements.txt
python init_federation.py {NR_OF_CLIENTS}

Creating a compute package

tar -czvf package/kerascifar.tar.gz client

Start the client

The easiest way to start clients for quick testing is by using Docker. We provide a docker-compose template for convenience. First, edit 'fedn-network.yaml' to provide information about the reducer endpoint. Then run following command in project directory: (Replace {NR_OF_CLIENTS} with the number of clients you selected)

docker-compose -f docker-compose-{NR_OF_CLIENTS}clients.yaml up 

Note that this assumes that a FEDn network is running with "keras" helper, which is identified in "config/settings-reducer.yaml" (see separate deployment instructions). The file 'docker-compose.dev.yaml' is for testing againts a local pseudo-distributed FEDn network. Use 'docker-compose.yaml' if you are connecting against a reducer part of a distributed setup and provide a 'extra_hosts' file.