:mod:`fedlab_core.client.topology`
==================================

.. py:module:: fedlab_core.client.topology


Module Contents
---------------

Classes
~~~~~~~

.. autoapisummary::

   fedlab_core.client.topology.ClientCommunicationTopology
   fedlab_core.client.topology.ClientSyncTop



.. class:: ClientCommunicationTopology(backend_handler, server_addr, world_size, rank, dist_backend)


   Bases: :class:`torch.multiprocessing.Process`

   Abstract class

   If you want to define your own Network Topology, please be sure your class should subclass it and OVERRIDE its
   methods.

   .. rubric:: Example

   please read the code of :class:`ClientSyncTop`

   .. method:: run(self)
      :abstractmethod:

      Please override this function


   .. method:: on_receive(self, sender, message_code, payload)
      :abstractmethod:

      Please override this function


   .. method:: synchronize(self)
      :abstractmethod:

      Please override this function



.. class:: ClientSyncTop(backend_handler, server_addr, world_size, rank, dist_backend='gloo', logger_file='clientLog', logger_name='')


   Bases: :class:`fedlab_core.client.topology.ClientCommunicationTopology`

   Synchronise communication class

   This is the top class in our framework which is mainly responsible for network communication of CLIENT!
   Synchronize with server following agreements defined in :meth:`run`.

   :param backend_handler: Subclass of ClientBackendHandler, manages training and evaluation of local model on each
   :param client.:
   :param server_addr: Address of server in form of ``(SERVER_ADDR, SERVER_IP)``
   :type server_addr: tuple
   :param world_size: Number of client processes participating in the job for ``torch.distributed`` initialization
   :type world_size: int
   :param rank: Rank of the current client process for ``torch.distributed`` initialization
   :type rank: int
   :param dist_backend: :attr:`backend` of ``torch.distributed``. Valid values include ``mpi``, ``gloo``,
   :type dist_backend: str or Backend
   :param and ``nccl``. Default: ``"gloo"``
   :param logger_file: Path to the log file for all clients of :class:`ClientSyncTop` class. Default: ``"clientLog"``
   :type logger_file: str, optional
   :param logger_name: Class name to initialize logger
   :type logger_name: str, optional

   :raises Errors raised by :func:`torch.distributed.init_process_group`:

   .. rubric:: Example

   TODO

   .. method:: run(self)

      Main procedure of each client is defined here:
      1. client waits for data from server
      2. after receiving data, client will train local model
      3. client will synchronize with server actively


   .. method:: on_receive(self, sender, message_code, payload)

      Actions to perform on receiving new message, including local training

      :param sender: Index of sender
      :type sender: int
      :param message_code: Agreements code defined in :class:`MessageCode` class
      :type message_code: MessageCode
      :param payload: Serialized model parameters
      :type payload: torch.Tensor


   .. method:: synchronize(self)

      Synchronize local model with server actively

      :param buffer: Serialized model parameters

      :returns: None

      :raises None:


   .. method:: _waiting(self)