Aggregating local models

The aggregation.py code handles aggregating local models with a bunch of aggregation algorithms. In the code example, we only support FedAvg, as discussed in the following sections.

Importing the libraries for the aggregator

The aggregation.py code imports the following:

import logging
import time
import numpy as np
from typing import List
from .state_manager import StateManager
from fl_main.lib.util.helpers import generate_model_id
from fl_main.lib.util.states import IDPrefix

The imported state manager’s role and functionalities are discussed in the Maintaining models for aggregation with the state manager section, and the helpers and states libraries are introduced in the Appendix, Exploring Internal Libraries.

After importing the necessary libraries, let’s define the aggregator class.

Defining and initializing the aggregator class

The following code for class Aggregator defines the core process of the aggregator, which...