fedlab_core.server.topology

Module Contents

Classes

EndTop

Abstract class for server network topology

ServerSyncTop

Synchronous communication class
class fedlab_core.server.topology.EndTop(server_handler, server_address, dist_backend)

Bases: torch.multiprocessing.Process

Abstract class for server network topology

If you want to define your own topology agreements, please subclass it.

Parameters

  • server_handler – Parameter server backend handler derived from ParameterServerHandler

  • server_address (tuple) – Address of server in form of (SERVER_ADDR, SERVER_IP)

  • dist_backend (str or Backend) – backend of torch.distributed. Valid values include mpi, gloo,

  • nccl (and) –

abstract run(self)

Process

abstract activate_clients(self)

Activate some of clients to join this FL round

abstract listen_clients(self)

Listen messages from clients

class fedlab_core.server.topology.ServerSyncTop(server_handler, server_address, dist_backend='gloo', logger_path='server_top.txt', logger_name='ServerTop')

Bases: fedlab_core.server.topology.EndTop

Synchronous communication class

This is the top class in our framework which is mainly responsible for network communication of SERVER!. Synchronize with clients following agreements defined in run().

Parameters

  • server_handler – Subclass of ParameterServerHandler

  • server_address (tuple) – Address of this server in form of (SERVER_ADDR, SERVER_IP)

  • dist_backend (str or Backend) – backend of torch.distributed. Valid values include mpi, gloo,

  • nccl. Default (and) – "gloo"

  • logger_path (str, optional) – path to the log file for this class. Default: "server_top.txt"

  • logger_name (str, optional) – class name to initialize logger. Default: "ServerTop"

Raises

None –

run(self)

Process

activate_clients(self)

Activate some of clients to join this FL round

listen_clients(self)

Listen messages from clients