fl4health.mixins.core_protocols module

class BasicClientProtocol(*args, **kwargs)

Bases: BasicClientProtocolPreSetup, Protocol

A minimal protocol for BasicClient focused on methods.

compute_evaluation_loss(preds, features, target)

Return type:

EvaluationLosses

compute_training_loss(preds, features, target)

Return type:

TrainingLosses

criterion: _Loss

initialize_all_model_weights(parameters, config)

Return type:

None

model: Module

optimizers: dict[str, Optimizer]

predict(input)

Return type:

tuple[dict[str, Tensor], dict[str, Tensor]]

test_loader: DataLoader | None

train_loader: DataLoader

transform_target(target)

Return type:

Tensor | dict[str, Tensor]

update_before_train(current_server_round)

Return type:

None

val_loader: DataLoader

validate(include_losses_in_metrics=False)

Return type:

tuple[float, dict[str, Union[bool, bytes, float, int, str]]]

class BasicClientProtocolPreSetup(*args, **kwargs)

Bases: NumPyClientMinimalProtocol, Protocol

A minimal protocol for BasicClient focused on methods.

compute_loss_and_additional_losses(preds, features, target)

Return type:

tuple[Tensor, dict[str, Tensor] | None]

device: device

get_criterion(config)

Return type:

_Loss

get_data_loaders(config)

Return type:

tuple[DataLoader, ...]

get_model(config)

Return type:

Module

get_optimizer(config)

Return type:

Optimizer | dict[str, Optimizer]

initialized: bool

setup_client(config)

Return type:

None

class NumPyClientMinimalProtocol(*args, **kwargs)

Bases: Protocol

A minimal protocol for NumPyClient with just essential methods.

evaluate(parameters, config)

Return type:

tuple[float, int, dict[str, Union[bool, bytes, float, int, str]]]

fit(parameters, config)

Return type:

tuple[list[ndarray[Any, dtype[Any]]], int, dict[str, Union[bool, bytes, float, int, str]]]

get_parameters(config)

Return type:

list[ndarray[Any, dtype[Any]]]

set_parameters(parameters, config, fitting_round)

Return type:

None

update_after_train(local_steps, loss_dict, config)

Return type:

None