🚦 Node State

NodeState centralizes critical information, enabling different components within a node to coordinate effectively during federated learning. Thread locks are used to maintain data consistency during concurrent operations like voting and aggregation.

Key Attributes

  • addr (str): The node’s address.

  • status (str): The node’s current status (e.g., “Idle”, “Learning”).

  • experiment (Experiment): The current experiment’s configuration.

  • simulation (bool): Whether the node is running in a simulation.

  • models_aggregated (Dict): Aggregated models received from neighbors.

  • nei_status (Dict): Neighbors’ current round numbers.

  • train_set (List): Nodes participating in the current training round.

  • train_set_votes (Dict): Votes for train set selection.

  • Thread Locks (various): Ensure thread-safe access to shared state.

Experiment

The Experiment class encapsulates the configuration and state of a federated learning experiment. It tracks the experiment’s name, the total number of rounds, and the current round.

from p2pfl.experiment import Experiment

experiment = Experiment(exp_name="my_experiment", total_rounds=10)
print(f"Experiment name: {experiment.exp_name}")
print(f"Total rounds: {experiment.total_rounds}")
print(f"Current round: {experiment.round}")

experiment.increase_round()
print(f"Current round after increase: {experiment.round}")