:mod:`fedlab_core.utils.messaging`
==================================

.. py:module:: fedlab_core.utils.messaging


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

Classes
~~~~~~~

.. autoapisummary::

   fedlab_core.utils.messaging.MessageCode



Functions
~~~~~~~~~

.. autoapisummary::

   fedlab_core.utils.messaging.send_message
   fedlab_core.utils.messaging.recv_message
   fedlab_core.utils.messaging.broadcast_message
   fedlab_core.utils.messaging.recv_broadcast_message


.. class:: MessageCode

   Bases: :class:`enum.Enum`

   Different types of messages between client and server that we support go here.

   .. attribute:: ParameterRequest
      :annotation: = 0

      

   .. attribute:: GradientUpdate
      :annotation: = 1

      

   .. attribute:: ParameterUpdate
      :annotation: = 2

      

   .. attribute:: EvaluateParams
      :annotation: = 3

      

   .. attribute:: Exit
      :annotation: = 4

      


.. function:: send_message(message_code, payload, dst=0)

   Sends a message to destination.

   Concatenates destination rank, message code and payload into a single tensor, then sends it.

   :param message_code: Type of message, defined in MessageCode
   :type message_code: MessageCode
   :param payload: Serialized tensor
   :param dst: Destination rank. Default is 0 which means the server
   :type dst: int, optional

   :returns: Serialized message package, including rank of current process rank, message code, and the payload


.. function:: recv_message(payload, src=None)

   Receives message from source.

   :param payload: Tensor to fill with received data. The first element is the source rank, and the second element is
   :param message code.:
   :param src: Source rank. Will receive from any process if unspecified.
   :type src: int, optional

   :returns: Serialized model parameters

   :raises None:


.. function:: broadcast_message(message_code, payload)

   Broadcast a message to all workers.

   Concatenates destination rank, message code and payload into a single tensor, then broadcasts the tensor to the
   whole group.

   Args:

   Returns:

   Raises:


.. function:: recv_broadcast_message(recv_buff)

   Workers recv the message from the center

   Args:

   Returns:

   Raises: