Base Metric Storage

Base Metric Storage#

Base abstract Metric Storage class

class rizemind.logging.base_metric_storage.BaseMetricStorage[source]#

Bases: ABC

An abstract base class (ABC) for metric storage backends.

This class defines the standard interface that all metric storage implementations must adhere to. It ensures that different storage mechanisms (e.g., in-memory, file-based, or a dedicated server) can be used interchangeably by a strategy that requires metric (i.e. MetricStorageStrategy).

To create a custom metric storage backend, you must subclass BaseMetricStorage and provide concrete implementations for all of its abstract methods.

abstractmethod update_best_model(server_round: int, loss: float) → None[source]#

Updates the stored “best” model if the current model’s loss is lower.

This method is responsible for tracking the best-performing model encountered throughout the entire federated learning process. It should compare the provided loss with a stored minimum loss and overwrite the saved “best model” if the new one is better.

Parameters:

  • server_round (int) – The server round that produced this model.

  • loss (float) – The loss value of the current model, used as the primary criterion for determining the “best” model.

abstractmethod update_current_round_model(parameters: Parameters) → None[source]#

Updates the model parameters from the most recent round.

This is used for temporarily keeping the parameters during the configure_evaluate phase so that it can later be written to the disk if the evaluation indicates it is the best model.

Parameters:

parameters (Parameters) – The model parameters from the current round to be saved.

abstractmethod write_metrics(server_round: int, metrics: dict[str, bool | bytes | float | int | str]) → None[source]#

Writes a dictionary of metrics for a specific server round.

This method is called to persist the metrics (e.g., accuracy, precision) during various phases.

Parameters:

  • server_round (int) – The current round of federated learning.

  • metrics (dict[str, Scalar]) – A dictionary mapping metric names (e.g., “accuracy”) to their scalar values.

Contents