First run

After installing fluke, you can run a federated algorithm from the command line. To do so, you need two configuration files:

• EXP_CONFIG_FILE: the experiment configuration file (independent from the algorithm);

• ALG_CONFIG_FILE: the algorithm configuration file;

Then, you can run the following command:

fluke --config=EXP_CONFIG_FILE federation ALG_CONFIG_FILE

Important

The fluke command provides also the possibility to run the algorithm in a centralized fashion as well as client-side only (no federation).

fluke --config=EXP_CONFIG_FILE centralized ALG_CONFIG_FILE
fluke --config=EXP_CONFIG_FILE clients-only ALG_CONFIG_FILE

In these settings, the same configuration files as the federated case can be used but some of the parameters might be ignored and some others are adjusted to be consistent with the selected setting. For example, the number of epochs in the centralized setting is the number of total rounds times the participation fraction, while in the clients-only setting it is the number of local epochs times the number of rounds times the participation fraction (anyway no less than 100 epochs). Please refer to this tutorial for more details.

For the purpose of this introductory page, we will use the configuration files provided in the configs folder of the fluke repository.

Tip

You can easily get the configuration files provided in the repository using the fluke-get command. For example, to get the configuration files used in this tutorial, you can run the following commands:

fluke-get exp
fluke-get fedavg

These two file will be downloaded in the “./config” folder.

In particular, we will use the configuration files exp.yaml and fedavg.yaml to run the classic FedAvg algorithm on the MNIST dataset.

fluke --config=path_to_configs/exp.yaml federation path_to_configs/fedavg.yaml

where path_to_configs is the path to the folder containing the configuration files.

For more details on the configuration files, you can refer to the Configuration section.