florist.api.clients.clients module¶

Implementation of the clients and the Client enumeration.

class Client(value)[source]¶

Bases: Enum

Enumeration of supported clients.

FEDAVG = 'FedAvg'¶

FEDPROX = 'FedProx'¶

get_client_class()[source]¶

Return the class for this client.

Return type:: type[LocalDataClient]
Returns:: (type[LocalDataClient]) A subclass of LocalDataClient corresponding to the this client.
Raises:: ValueError – if the client is not supported.

classmethod list()[source]¶

List all the supported clients.

Return type:: list[str]
Returns:: (list[str]) a list of supported clients.

classmethod list_by_strategy(strategy)[source]¶

List all the supported clients given a strategy.

Parameters:: strategy (Strategy) – (Strategy) the strategy to find the supported clients.
Return type:: List[str]
Returns:: (list[str]) a list of supported clients for the given strategy.

class FedProxLocalDataClient(data_path, metrics, device, loss_meter_type=LossMeterType.AVERAGE, checkpoint_and_state_module=None, reporters=None, progress_bar=False, client_name=None)[source]¶

Bases: FedProxClient, LocalDataClient

Implementation of the FedProx client that uses a model with data stored locally.

get_data_loaders(config)[source]¶

Return the data loader for FedProx on model with data stored locally.

Parameters:: config (dict[str, Union[bool, bytes, float, int, str]]) – (Config) the Config object for this client.
Return type:: tuple[DataLoader[TensorDataset], DataLoader[TensorDataset]]
Returns:: (Tuple[DataLoader[TensorDataset], DataLoader[TensorDataset]]) a tuple with the train data loader and validation data loader respectively.

class LocalDataClient(data_path, metrics, device, loss_meter_type=LossMeterType.AVERAGE, checkpoint_and_state_module=None, reporters=None, progress_bar=False, client_name=None)[source]¶

Bases: BasicClient

Implementation of a client that uses a model with data stored locally.

get_criterion(config)[source]¶

Return the loss for the model.

Parameters:: config (dict[str, Union[bool, bytes, float, int, str]]) – (Config) the Config object for this client.
Return type:: _Loss
Returns:: (torch.nn.modules.loss._Loss) an instance of torch.nn.modules.loss._Loss that has been defined by the local model.

get_data_loaders(config)[source]¶

Return the data loader for the model with local data.

Parameters:: config (dict[str, Union[bool, bytes, float, int, str]]) – (Config) the Config object for this client.
Return type:: tuple[DataLoader[TensorDataset], DataLoader[TensorDataset]]
Returns:: (Tuple[DataLoader[TensorDataset], DataLoader[TensorDataset]]) a tuple with the train data loader and validation data loader respectively.

get_model(config)[source]¶

Return the model for training with local data.

Parameters:: config (dict[str, Union[bool, bytes, float, int, str]]) – (Config) the Config object for this client.
Return type:: Module
Returns:: (torch.nn.Module) An instance of the model.

get_optimizer(config)[source]¶

Return the optimizer for the model.

Parameters:: config (dict[str, Union[bool, bytes, float, int, str]]) – (Config) the Config object for this client.
Return type:: Optimizer
Returns:: (torch.optim.Optimizer) An instance of torch.optim.Optimizer with the configurations defined by self.optimizer_type.

set_model(model)[source]¶

Set the model to be used for training with local data.

Parameters:: model (LocalDataModel) – (LocalModel) An instance of the model to be used for training.
Return type:: None

set_optimizer_type(optimizer_type)[source]¶

Set the type of the optimizer to be used for training.

Parameters:: optimizer_type (Optimizer) – (Optimizer) A value of the Optimizer enumeration with the type of the optimizer to be used for training.
Return type:: None