torch/distributed/rpc/api.py

import collections
import contextlib
import functools
import logging
import numbers
import sys
import threading
from datetime import timedelta

import torch
import torch.distributed as dist

from . import (
    RpcBackendOptions,
    WorkerInfo,
    _cleanup_python_rpc_handler,
    _delete_all_user_rrefs,
    _destroy_rref_context,
    _get_current_rpc_agent,
    _invoke_remote_builtin,
    _invoke_remote_python_udf,
    _invoke_remote_torchscript,
    _invoke_rpc_builtin,
    _invoke_rpc_python_udf,
    _invoke_rpc_torchscript,
    _is_current_rpc_agent_set,
    _reset_current_rpc_agent,
    _set_and_start_rpc_agent,
    _set_rpc_timeout,
    backend_registry,
)
from .internal import (
    PythonUDF,
    RPCExecMode,
    _internal_rpc_pickler,
    _start_record_function,
)


logger = logging.getLogger(__name__)


# NB: Ignoring RRef leaks during shutdown. Without this, applications have to
# make sure there is no references to any RRef in the application code and
# Python GC has done its job to delete those RRefs. This is could result in bad
# debugging experiences especially when for large applications. Therefore, by
# default, we are going to ignore RRef leaks during shutdown. This is usually
# fine as shutdown means applications have done training and no longer care
# about states.
#
# To enable RRef leak checking, set this _ignore_rref_leak to False
_ignore_rref_leak = True
_default_pickler = _internal_rpc_pickler


@contextlib.contextmanager
def _use_rpc_pickler(rpc_pickler):
    r"""
    rpc_pickler: (.internal._InternalRPCPickler) Overrides the default RPC pickler
    """
    global _default_pickler
    _default_pickler = rpc_pickler
    try:
        yield
    finally:
        _default_pickler = _internal_rpc_pickler


def _require_initialized(func):
    @functools.wraps(func)
    def wrapper(*args, **kwargs):
        if not _is_current_rpc_agent_set():
            raise RuntimeError(
                "RPC has not been initialized. Call "
                "torch.distributed.rpc.init_rpc first."
            )
        return func(*args, **kwargs)

    return wrapper


class WaitAllWorkersStates(object):
    def __init__(self):
        # Each `intent_worker_names` is an empty set at beginning.
        # It's only used by leader worker. Leader worker is user-specified or
        # elected as the first worker in a sorted worker name list.
        # Whenever there is a worker showing shutdown intention to the leader, by
        # calling `_wait_all_workers()`, the leader adds this worker's name to the set.
        # The leader also adds itself's name to the set on calling
        # `_wait_all_workers()`. We need this because, we confine `_wait_all_workers()`
        # to be called only once, by examing if leader's name has been added to the set.
        self.intent_worker_names = set()
        # Once `intent_worker_names == _ALL_WORKER_NAMES`,
        # we flip `_SHUTDOWN_PROCEED_SIGNAL` on the leader, and leader will send RPCs
        # to follower workers to flip their `_SHUTDOWN_PROCEED_SIGNAL`s.
        self.proceed_signal = threading.Event()


# States used by `def _wait_all_workers()`.
# `_ALL_WORKER_NAMES` is initialized on initiaizing RPC layer.
_ALL_WORKER_NAMES = None
_wait_all_workers_dict_lock = threading.Lock()
_wait_all_workers_sequence_id = 0
_wait_all_workers_sequence_id_to_states = collections.defaultdict(WaitAllWorkersStates)


def _on_leader_follower_report_shutdown_intent(sequence_id, worker_name):
    assert (
        worker_name in _ALL_WORKER_NAMES
    ), "{worker_name} is not expected by leader.".format(worker_name=worker_name)
    intent_worker_names = _wait_all_workers_sequence_id_to_states[
        sequence_id
    ].intent_worker_names
    assert (
        worker_name not in intent_worker_names
    ), "{worker_name} reported intent sequence id {sequence_id} twice. ".format(
        worker_name=worker_name, sequence_id=sequence_id
    )
    intent_worker_names.add(worker_name)
    if _ALL_WORKER_NAMES == intent_worker_names:
        _set_proceed_shutdown_signal(sequence_id)


def _set_proceed_shutdown_signal(sequence_id):
    proceed_signal = _wait_all_workers_sequence_id_to_states[sequence_id].proceed_signal
    assert (
        not proceed_signal.is_set()
    ), "Termination signal sequence id {} got set twice.".format(
        sequence_id=sequence_id
    )
    proceed_signal.set()


@_require_initialized
def _wait_all_workers():
    r"""
    Block until all local and remote RPC processes reach this method and wait
    for all outstanding work to complete. Every RPC process must call this
    method before exit to perform a graceful shutdown. This should be used to
    terminate the RPC framework, and there is no guarantee that the RPC
    framework will work after this method returns.
    """
    assert (
        _ALL_WORKER_NAMES is not None
    ), "`_ALL_WORKER_NAMES` is not initialized for `def _wait_all_workers`."
    leader_worker_name = sorted(_ALL_WORKER_NAMES)[0]

    self_worker_name = _get_current_rpc_agent().get_worker_info().name

    global _wait_all_workers_sequence_id
    with _wait_all_workers_dict_lock:
        sequence_id = _wait_all_workers_sequence_id
        _wait_all_workers_sequence_id += 1

    is_leader_worker = leader_worker_name == self_worker_name

    # Phase 1: Followers send intents.
    # All followers report intents to the leader.
    if is_leader_worker:
        _on_leader_follower_report_shutdown_intent(sequence_id, self_worker_name)
    else:
        rpc_sync(
            leader_worker_name,
            _on_leader_follower_report_shutdown_intent,
            args=(sequence_id, self_worker_name,),
        )

    proceed_signal = _wait_all_workers_sequence_id_to_states[
        sequence_id
    ].proceed_signal
    proceed_signal.wait()

    # Phase 2: Leader asks followers to proceed.
    # Leader's signal is the first to be unblocked,
    # after receiving all followers' intents.
    if is_leader_worker:
        # The leader sends out proceeed signals to all followers.
        timeout = timedelta(seconds=5)
        _set_rpc_timeout(timeout)
        worker_name_to_response_future_dict = dict()
        for follower_worker_name in _ALL_WORKER_NAMES - {leader_worker_name}:
            fut = rpc_async(follower_worker_name, _set_proceed_shutdown_signal, args=(sequence_id,))
            worker_name_to_response_future_dict[follower_worker_name] = fut
        for follower_worker_name, fut in worker_name_to_response_future_dict.items():
            try:
                fut.wait()
            except RuntimeError as ex:
                logger.error(
                    "{worker_name} failed to respond to 'Shutdown Proceed.' request in {timeout}".format(
                        worker_name=follower_worker_name, timeout=timeout
                    )
                )


@_require_initialized
def shutdown(graceful=True):
    r"""
    Perform a shutdown of the RPC agent, and then destroy the RPC agent. This
    stops the local agent from  accepting outstanding requests, and shuts
    down the RPC framework by terminating all RPC threads. If graceful=True,
    then this will block until all local and remote RPC processes reach this
    method and wait for all outstanding work to complete. Otherwise, if
    graceful=False, then this is a local shutdown, and it does not wait for
    other RPC processes to reach this method.

    Arguments:
        graceful (bool): Whether to do a graceful shutdown or not. If True,
                         this will 1) wait until there is no pending system
                         messages for ``UserRRef``s and delete them; 2) block
                         until all local and remote RPC processes have reached
                         this method and wait for all outstanding work to
                         complete.

    Example::
        Make sure that ``MASTER_ADDRESS`` and ``MASTER_PORT`` are set properly
        on both workers. Refer to :meth:`~torch.distributed.init_process_group`
        API for more details. For example,

        >>> export MASTER_ADDRESS=localhost
        >>> export MASTER_port=5678

        Then run the following code in two different processes:

        >>> # On worker 0:
        >>> import torch
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker0", rank=0, world_size=2)
        >>> # do some work
        >>> result = rpc.rpc_sync("worker1", torch.add, args=(torch.ones(1), 1))
        >>> # ready to shutdown
        >>> rpc.shutdown()

        >>> # On worker 1:
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker1", rank=1, world_size=2)
        >>> # wait for worker 0 to finish work, and then shutdown.
        >>> rpc.shutdown()
    """
    if graceful:
        _wait_all_workers()
        _delete_all_user_rrefs()
        _get_current_rpc_agent().join()
    try:
        # This raises a `TORCH_CHECK()` exception on RRef leak detected.
        _destroy_rref_context(_ignore_rref_leak)
    finally:
        _get_current_rpc_agent().shutdown()
        # clean up python rpc handler in shutdown(), see comments in
        # PythonRpcHandler::cleanup(), call it in python API because the
        # cleanup() function has python dependency, it assumes python
        # interpreter exists.
        # No matter if RRef leak exception is raised, this clean-up code
        # must run to avoid destruction segfault in Python 3.5.
        _cleanup_python_rpc_handler()
        _reset_current_rpc_agent()


# TODO: add a context manager to wrap _init_rpc_backend and shutdown
def _init_rpc_backend(
    backend=backend_registry.BackendType.PROCESS_GROUP,
    store=None,
    name=None,
    rank=-1,
    world_size=-1,
    rpc_backend_options=None,
):

    if sys.version_info < (3, 0):
        raise RuntimeError("RPC package does not support Python2.")

    _validate_rpc_args(backend, store, name, rank, world_size, rpc_backend_options)

    if _is_current_rpc_agent_set():
        raise RuntimeError("RPC is already initialized")

    # Initialize RPC.
    rpc_agent = backend_registry.init_backend(
        backend,
        store=store,
        name=name,
        rank=rank,
        world_size=world_size,
        rpc_backend_options=rpc_backend_options,
    )

    worker_infos = rpc_agent.get_worker_infos()
    global _ALL_WORKER_NAMES
    _ALL_WORKER_NAMES = {worker_info.name for worker_info in worker_infos}

    _set_and_start_rpc_agent(rpc_agent)


@_require_initialized
def get_worker_info(worker_name=None):
    r"""
    Get :class:`~torch.distributed.rpc.WorkerInfo` of a given worker name.
    Use this :class:`~torch.distributed.rpc.WorkerInfo` to avoid passing an
    expensive string on every invocation.

    Arguments:
        worker_name (str): the string name of a worker. If ``None``, return the
                           the id of the current worker. (default ``None``)

    Returns:
        :class:`~torch.distributed.rpc.WorkerInfo` instance for the given
        ``worker_name`` or :class:`~torch.distributed.rpc.WorkerInfo` of the
        current worker if ``worker_name`` is ``None``.
    """
    if worker_name:
        return _get_current_rpc_agent().get_worker_info(worker_name)
    else:
        return _get_current_rpc_agent().get_worker_info()


def _to_worker_info(name_or_info):
    if isinstance(name_or_info, WorkerInfo):
        return name_or_info
    elif isinstance(name_or_info, str):
        return get_worker_info(name_or_info)
    else:
        raise ValueError("Cannot get WorkerInfo from name {}".format(name_or_info))


def _validate_rpc_args(backend, store, name, rank, world_size, rpc_backend_options):
    type_mapping = {
        backend: backend_registry.BackendType,
        store: dist.Store,
        name: str,
        rank: numbers.Integral,
        world_size: numbers.Integral,
        rpc_backend_options: RpcBackendOptions,
    }
    for arg, arg_type in type_mapping.items():
        if not isinstance(arg, arg_type):
            raise RuntimeError(
                "Argument {} must be of type {} but got type {}".format(
                    arg, arg_type, type(arg)
                )
            )


@_require_initialized
def remote(to, func, args=None, kwargs=None):
    r"""
    Make a remote call to run ``func`` on worker ``to`` and return an
    :class:`~torch.distributed.rpc.RRef` to the result value immediately.
    Worker ``to`` will be the owner of the returned
    :class:`~torch.distributed.rpc.RRef`, and the worker calling ``remote`` is
    a user. The owner manages the global reference count of its
    :class:`~torch.distributed.rpc.RRef`, and the owner
    :class:`~torch.distributed.rpc.RRef` is only destructed when globally there
    are no living references to it.

    Arguments:
        to (str or WorkerInfo): id or name of the destination worker.
        func (callable): any callable function. python callable, builtin or annotated TorchScript
                         functions (like meth:`torch.add`) can be sent over RPC more efficiently.
        args (tuple): the argument tuple for the ``func`` invocation.
        kwargs (dict): is a dictionary of keyword arguments for the ``func``
                       invocation.

    Returns:
        A user :class:`~torch.distributed.rpc.RRef` instance to the result
        value. Use the blocking API :meth:`torch.distributed.rpc.RRef.to_here`
        to retrieve the result value locally.

    .. warning ::
        Using GPU tensors as arguments or return values of ``func`` is not
        supported since we don't support sending GPU tensors over the wire. You
        need to explicitly copy GPU tensors to CPU before using them as
        arguments or return values of ``func``.

    Example::
        Make sure that ``MASTER_ADDRESS`` and ``MASTER_PORT`` are set properly
        on both workers. Refer to :meth:`~torch.distributed.init_process_group`
        API for more details. For example,

        >>> export MASTER_ADDRESS=localhost
        >>> export MASTER_port=5678

        Then run the following code in two different processes:

        >>> # On worker 0:
        >>> import torch
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker0", rank=0, world_size=2)
        >>> rref1 = rpc.remote("worker1", torch.add, args=(torch.ones(2), 3))
        >>> rref2 = rpc.remote("worker1", torch.add, args=(torch.ones(2), 1))
        >>> x = rref1.to_here() + rref2.to_here()
        >>> rpc.shutdown()

        >>> # On worker 1:
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker1", rank=1, world_size=2)
        >>> rpc.shutdown()

        If invoking an annotated TorchScript function, then run the following
        code in two different processes:

        >>> # On worker 0:
        >>> @torch.jit.script
        >>> def my_script_add(t1, t2):
        >>>    return torch.add(t1, t2)
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker0", rank=0, world_size=2)
        >>> rref = rpc.remote("worker1", my_script_add, args=(torch.ones(2), 3))
        >>> rref.to_here()
        >>> rpc.shutdown()

        >>> # On worker 1:
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker1", rank=1, world_size=2)
        >>> rpc.shutdown()
    """
    qualified_name = torch.jit._find_builtin(func)
    dst_worker_info = _to_worker_info(to)

    # If profiling is enabled, kick off the timer and retrieve back a
    # RecordFunction instance.
    rf = None
    if torch.autograd._profiler_enabled():
        rf = _start_record_function(
            RPCExecMode.REMOTE,
            str(qualified_name) if qualified_name is not None else func.__qualname__,
            get_worker_info().name,
            dst_worker_info.name,
        )

    args = args if args else ()
    kwargs = kwargs if kwargs else {}

    if qualified_name is not None:
        return _invoke_remote_builtin(dst_worker_info, qualified_name, rf, *args, **kwargs)
    elif isinstance(func, torch.jit.ScriptFunction):
        return _invoke_remote_torchscript(
            dst_worker_info.name, torch._jit_internal._qualified_name(func), *args, **kwargs
        )
    else:
        (pickled_python_udf, tensors) = _default_pickler.serialize(
            PythonUDF(func, args, kwargs)
        )
        return _invoke_remote_python_udf(dst_worker_info, pickled_python_udf, tensors, rf)


def _invoke_rpc(to, func, rpc_type, args=None, kwargs=None):
    if not callable(func):
        raise TypeError("function should be callable.")

    qualified_name = torch.jit._find_builtin(func)
    dst_worker_info = _to_worker_info(to)
    # If profiling is enabled, kick off the timer and retrieve back a
    # RecordFunction instance.
    rf = None
    if torch.autograd._profiler_enabled():
        rf = _start_record_function(
            rpc_type,
            str(qualified_name) if qualified_name is not None else func.__qualname__,
            get_worker_info().name,
            dst_worker_info.name,
        )

    args = args if args else ()
    kwargs = kwargs if kwargs else {}

    if qualified_name is not None:
        fut = _invoke_rpc_builtin(dst_worker_info, qualified_name, rf, *args, **kwargs)
    elif isinstance(func, torch.jit.ScriptFunction):
        fut = _invoke_rpc_torchscript(
            dst_worker_info.name, torch.jit._qualified_name(func), *args, **kwargs
        )
    else:
        (pickled_python_udf, tensors) = _default_pickler.serialize(
            PythonUDF(func, args, kwargs)
        )
        fut = _invoke_rpc_python_udf(dst_worker_info, pickled_python_udf, tensors, rf)
    return fut


@_require_initialized
def rpc_sync(to, func, args=None, kwargs=None):
    r"""
    Make a blocking RPC call to run function ``func`` on worker ``to``. RPC
    messages are sent and received in parallel to execution of Python code. This
    method is thread-safe.

    Arguments:
        to (str or WorkerInfo): id or name of the destination worker.
        func (callable): any callable function. python callable, builtin or annotated TorchScript
                         functions (like meth:`torch.add`) can be sent over RPC more efficiently.
        args (tuple): the argument tuple for the ``func`` invocation.
        kwargs (dict): is a dictionary of keyword arguments for the ``func``
                       invocation.

    Returns:
        Returns the result of running ``func`` on ``args`` and ``kwargs``.

    .. warning ::
        Using GPU tensors as arguments or return values of ``func`` is not
        supported since we don't support sending GPU tensors over the wire. You
        need to explicitly copy GPU tensors to CPU before using them as
        arguments or return values of ``func``.

    Example::
        Make sure that ``MASTER_ADDRESS`` and ``MASTER_PORT`` are set properly
        on both workers. Refer to :meth:`~torch.distributed.init_process_group`
        API for more details. For example,

        >>> export MASTER_ADDRESS=localhost
        >>> export MASTER_port=5678

        Then run the following code in two different processes:

        >>> # On worker 0:
        >>> import torch
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker0", rank=0, world_size=2)
        >>> ret = rpc.rpc_sync("worker1", torch.add, args=(torch.ones(2), 3))
        >>> rpc.shutdown()

        >>> # On worker 1:
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker1", rank=1, world_size=2)
        >>> rpc.shutdown()

        If invoking an annotated TorchScript function, then run the following
        code in two different processes:

        >>> # On worker 0:
        >>> @torch.jit.script
        >>> def my_script_add(t1, t2):
        >>>    return torch.add(t1, t2)
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker0", rank=0, world_size=2)
        >>> ret = rpc.rpc_sync("worker1", my_script_add, args=(torch.ones(2), 3))
        >>> rpc.shutdown()

        >>> # On worker 1:
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker1", rank=1, world_size=2)
        >>> rpc.shutdown()

    """
    fut = _invoke_rpc(to, func, RPCExecMode.SYNC, args, kwargs)
    return fut.wait()


@_require_initialized
def rpc_async(to, func, args=None, kwargs=None):
    r"""
    Make a non-blocking RPC call to run function ``func`` on worker ``to``. RPC
    messages are sent and received in parallel to execution of Python code. This
    method is thread-safe. This method will immediately return a
    Future that can be awaited on.

    Arguments:
        to (str or WorkerInfo): id or name of the destination worker.
        func (callable): any callable function. python callable, builtin or annotated TorchScript
                         functions (like meth:`torch.add`) can be sent over RPC more efficiently.
        args (tuple): the argument tuple for the ``func`` invocation.
        kwargs (dict): is a dictionary of keyword arguments for the ``func``
                       invocation.

    Returns:
        Returns a Future object that can be waited
        on. When completed, the return value of ``func`` on ``args`` and
        ``kwargs`` can be retrieved from the Future object.

    .. warning ::
        Using GPU tensors as arguments or return values of ``func`` is not
        supported since we don't support sending GPU tensors over the wire. You
        need to explicitly copy GPU tensors to CPU before using them as
        arguments or return values of ``func``.

    Example::
        Make sure that ``MASTER_ADDRESS`` and ``MASTER_PORT`` are set properly
        on both workers. Refer to :meth:`~torch.distributed.init_process_group`
        API for more details. For example,

        >>> export MASTER_ADDRESS=localhost
        >>> export MASTER_port=5678

        Then run the following code in two different processes:

        >>> # On worker 0:
        >>> import torch
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker0", rank=0, world_size=2)
        >>> fut1 = rpc.rpc_async("worker1", torch.add, args=(torch.ones(2), 3))
        >>> fut2 = rpc.rpc_async("worker1", min, args=(1, 2))
        >>> result = fut1.wait() + fut2.wait()
        >>> rpc.shutdown()

        >>> # On worker 1:
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker1", rank=1, world_size=2)
        >>> rpc.shutdown()

        If invoking an annotated TorchScript function, then run the following
        code in two different processes:

        >>> # On worker 0:
        >>> @torch.jit.script
        >>> def my_script_add(t1, t2):
        >>>    return torch.add(t1, t2)
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker0", rank=0, world_size=2)
        >>> fut = rpc.rpc_async("worker1", my_script_add, args=(torch.ones(2), 3))
        >>> ret = fut.wait()
        >>> rpc.shutdown()

        >>> # On worker 1:
        >>> import torch.distributed.rpc as rpc
        >>> rpc.init_rpc("worker1", rank=1, world_size=2)
        >>> rpc.shutdown()
    """
    return _invoke_rpc(to, func, RPCExecMode.ASYNC, args, kwargs)