# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. import numpy as np import os import pickle import io from datetime import timedelta from ..fluid.layer_helper import LayerHelper from ..fluid.framework import Variable from ..fluid.framework import in_dygraph_mode from ..fluid.framework import OpProtoHolder from ..fluid.framework import _non_static_mode from ..fluid.framework import _in_legacy_dygraph from ..fluid.framework import convert_np_dtype_to_dtype_ from ..fluid.framework import _varbase_creator from ..fluid.data_feeder import convert_dtype from ..fluid.data_feeder import check_variable_and_dtype from ..fluid.data_feeder import check_type from ..fluid.data_feeder import check_dtype from ..fluid.layers.tensor import fill_constant from ..fluid.layers import utils from ..fluid.dygraph import layers from ..fluid.dygraph.parallel import prepare_context import paddle import paddle.fluid as fluid import paddle.fluid.core as core from paddle import _C_ops import paddle.fluid.dygraph_utils as dygraph_utils import contextlib __all__ = [] class ReduceOp: """ Specify the type of operation used for element-wise reductions. It should be one of the following values: ReduceOp.SUM ReduceOp.MAX ReduceOp.MIN ReduceOp.PROD Examples: .. code-block:: python import numpy as np import paddle from paddle.distributed import ReduceOp from paddle.distributed import init_parallel_env paddle.set_device('gpu:%d'%paddle.distributed.ParallelEnv().dev_id) init_parallel_env() if paddle.distributed.ParallelEnv().local_rank == 0: np_data = np.array([[4, 5, 6], [4, 5, 6]]) else: np_data = np.array([[1, 2, 3], [1, 2, 3]]) data = paddle.to_tensor(np_data) paddle.distributed.all_reduce(data, op=ReduceOp.SUM) out = data.numpy() # [[5, 7, 9], [5, 7, 9]] """ SUM = 0 MAX = 1 MIN = 2 PROD = 3 AVG = 4 class Group(): """ The abstract representation of group. """ def __init__(self, rank, rank_num, id=0, ranks=[], pg=None, name=None): self.rank = rank self.nranks = rank_num self.id = id self.ranks = ranks self.pg = pg self.name = name def is_member(self): if self.rank < 0: return False if self.nranks < 2: return False return True def get_group_rank(self, rank): if self.is_member() and rank in self.ranks: return self.ranks.index(rank) else: return -1 @property def process_group(self): return self.pg def __repr__(self): debug_str = "rank: {}, nranks: {}, id: {}, ranks: ".format( self.rank, self.nranks, self.id) debug_str += ", ".join(map(str, self.ranks)) debug_str += "; name: " debug_str += self.name if self.name else "None" return debug_str _global_env = None def _get_global_env(): global _global_env if not _global_env: _global_env = paddle.distributed.ParallelEnv() return _global_env # group map : the map of all group, 0 for GlobalGroup # Dict[int, Group] _group_map = {} # group map by name : the map of all groups from their names # Dict[name, Group] _group_map_by_name = {} # backend map by group : the map of all backend from their groups # Dict[group, backend] _group_map_backend = {} # Name of the default group for init_parallel_env _default_group_name = "_default_pg" _valid_backend_list = ['nccl', 'gloo', 'hccl', 'heter', 'xccl'] _default_store = None # the default tcp store _default_backend = None def _set_default_backend(backend): global _default_backend _default_backend = backend def _set_default_store(store): global _default_store _default_store = store def _get_group_map(): global _group_map if not _group_map: genv = _get_global_env() _group_map[0] = Group(genv.rank, genv.world_size, ranks=list(range(genv.world_size))) return _group_map def _get_global_group(): return _get_group_map()[0] def _get_group_map_by_name(): global _group_map_by_name return _group_map_by_name def _get_default_group(): global _group_map_by_name assert is_initialized(), ("Call paddle.distributed.init_parallel_env first " "to initialize the distributed environment.") return _get_group_map_by_name()[_default_group_name] def _set_group_map(gid, group): global _group_map assert gid not in _group_map _group_map[gid] = group def _set_group_map_by_name(name, group): global _group_map_by_name assert name not in _group_map_by_name _group_map_by_name[name] = group def _set_group_map_backend(group, backend): global _group_map_backend assert group not in _group_map_backend _group_map_backend[group] = backend def _new_ring_id(): return len(_get_group_map()) + max(_get_global_env().nrings, 9) def _get_reduce_op(reduce_op, func_name): if reduce_op == ReduceOp.SUM: return core.ReduceOp.SUM elif reduce_op == ReduceOp.MAX: return core.ReduceOp.MAX elif reduce_op == ReduceOp.MIN: return core.ReduceOp.MIN elif reduce_op == ReduceOp.PROD: return core.ReduceOp.PRODUCT else: raise ValueError("Unknown reduce_op type for {}.".format(func_name)) def get_group(id=0): """ Get group instance by group id. Args: id (int): the group id. Default value is 0. Returns: Group: the group instance. Examples: .. code-block:: python ... gid = paddle.distributed.new_group([2,4,6]) paddle.distributed.get_group(gid.id) """ gm = _get_group_map() return gm[id] if id in gm else None def _new_process_group_impl(backend, store, rank, world_size, group_name, pg_options, group_id=0, src_rank=None, dst_rank=None): pg = None genv = _get_global_env() if backend != 'heter': assert src_rank is None and dst_rank is None, ( "src_rank and dst_rank " "can only be set for heter backend.") assert backend in _valid_backend_list, "Unsupported backend: %s." % backend if backend == "gloo": place = core.CPUPlace() pg = core.ProcessGroupGloo(store, rank, world_size, place, group_id) elif backend == "nccl": place = core.CUDAPlace(genv.device_id) pg = core.ProcessGroupNCCL(store, rank, world_size, place, group_id) elif backend == "hccl": place = core.NPUPlace(genv.device_id) pg = core.ProcessGroupHCCL(store, rank, world_size, place, group_id) elif backend == "xccl": place = core.CustomPlace(genv.device_type, genv.device_id) pg = core.ProcessGroupCustom(store, rank, world_size, place, group_id) elif backend == "heter": place = None if core.is_compiled_with_cuda(): place = core.CUDAPlace(genv.device_id) elif core.is_compiled_with_npu(): place = core.NPUPlace(genv.device_id) cluster_id = int(os.getenv("CLUSTER_ID", "-1")) assert cluster_id >= 0, "please set the CLUSTER_ID variable." cluster_size = os.getenv("CLUSTER_SIZE", None) assert cluster_size, "please set the CLUSTER_SIZE variable." cluster_size = cluster_size.split(",") cluster_size = [int(s) for s in cluster_size] switch_ep = os.getenv("CLUSTER_SWITCH", None) assert switch_ep, "please set the CLUSTER_SWITCH variable." cluster_size_cumsum = np.cumsum(cluster_size) cluster_offset = 0 if cluster_id == 0 else cluster_size_cumsum[ cluster_id - 1] global_rank = cluster_offset + rank global_world_size = cluster_size_cumsum[-1] pg = core.ProcessGroupHeter(store, rank=global_rank, world_size=global_world_size, place=place, gid=group_id, local_rank=rank, local_size=world_size, gloo_rank=cluster_id, gloo_size=len(cluster_size), with_switch=True, switch_endpoint=switch_ep, src_rank=src_rank, dst_rank=dst_rank) return pg def barrier(group=None): """ Barrier among all participators in the group. Args: group (Group): The group instance return by new_group or None for global default group. Returns: None. Examples: .. code-block:: python import paddle from paddle.distributed import init_parallel_env paddle.set_device('gpu:%d'%paddle.distributed.ParallelEnv().dev_id) init_parallel_env() paddle.distributed.barrier() """ if group is not None and not group.is_member(): return if in_dygraph_mode(): group = _get_default_group() if group is None else group task = group.process_group.barrier() task.wait() return ring_id = 0 if group is None else group.id temp = fill_constant([1], dtype="int32", value="1") if _non_static_mode(): return _C_ops.barrier(temp, temp, 'ring_id', ring_id) op_type = 'barrier' if not isinstance(ring_id, int): raise ValueError("The type of 'group' for barrier must be int.") helper = LayerHelper(op_type, **locals()) helper.append_op(type=op_type, inputs={'X': [temp]}, outputs={'Out': [temp]}, attrs={'ring_id': ring_id}) # _custom_gid provides a way for users to # set the group id, which is usually useful # to be compatible with the static mode. _custom_gid = None def _set_custom_gid(gid): global _custom_gid _custom_gid = gid def new_group(ranks=None, backend=None): """ Creates a new distributed communication group. Args: ranks (list): The global ranks of group members. backend (str): The backend used to create group, only nccl is supported now. Returns: Group: The group instance. Examples: .. code-block:: python import paddle paddle.distributed.init_parallel_env() tindata = paddle.randn(shape=[2, 3]) gp = paddle.distributed.new_group([2,4,6]) paddle.distributed.all_reduce(tindata, group=gp, use_calc_stream=False) """ global _custom_gid global _group_map if in_dygraph_mode(): global _default_group_name gid = _custom_gid if _custom_gid else _new_ring_id() group_name = _default_group_name + str(gid) if backend != 'heter' and (ranks is None or len(ranks) > 1): global_group = _get_default_group() global_rank = global_group.rank global_ranks = global_group.ranks backend = _default_backend if backend is None else backend if ranks is None: ranks = global_ranks assert len(ranks) <= len(global_ranks), ( "Size of new group must be less than or " "equal to that of the default global group.") size = len(ranks) ranks = sorted(ranks) if backend == 'heter' or (size > 1 and global_rank in ranks): rank = 0 if backend == 'heter' else ranks.index(global_rank) src_rank = ranks[0] if backend == 'heter' else None dst_rank = ranks[1] if backend == 'heter' else None pg = _new_process_group_impl(backend, _default_store, rank, size, group_name, pg_options=None, group_id=gid, src_rank=src_rank, dst_rank=dst_rank) else: rank = -1 pg = None group = Group(rank, size, id=gid, ranks=ranks, pg=pg, name=group_name) _group_map_by_name[group_name] = group _group_map[gid] = group _group_map_backend[group] = backend # TODO(shenliang03): This is a temporary solution to solve the problem of # hang caused by tcp paddle.distributed.barrier(group=group) return group if not backend: backend = 'nccl' assert backend == 'nccl', ("backend other than nccl is not supported yet") genv = _get_global_env() global_rank = genv.rank ring_id = _new_ring_id() if global_rank not in ranks: gp = Group(-1, -1, ring_id, ranks) _group_map[ring_id] = gp else: ranks = sorted(ranks) group_rank = ranks.index(global_rank) group_size = len(ranks) gp = Group(group_rank, group_size, ring_id, ranks) _group_map[ring_id] = gp if group_size >= 2: strategy = core.ParallelStrategy() strategy.nranks = group_size strategy.local_rank = group_rank strategy.trainer_endpoints = [ genv.trainer_endpoints[i] for i in ranks ] strategy.current_endpoint = genv.current_endpoint strategy.nrings = 1 if core.is_compiled_with_cuda(): place = core.CUDAPlace(genv.device_id) core.NCCLParallelContext(strategy, place).init_with_ring_id(ring_id) elif core.is_compiled_with_npu(): place = core.NPUPlace(genv.device_id) core.HCCLParallelContext(strategy, place).init_with_ring_id(ring_id) elif core.is_compiled_with_mlu(): place = core.MLUPlace(genv.device_id) core.CNCLParallelContext(strategy, place).init_with_ring_id(ring_id) elif core.is_compiled_with_xpu(): place = core.XPUPlace(genv.device_id) core.BKCLParallelContext(strategy, place).init_with_ring_id(ring_id) else: assert False, ("no cuda device found") else: return gp # TODO(shenliang03): This is a temporary solution to solve the problem of # hang caused by cross-creation of new_group tmp = paddle.to_tensor( [1], dtype="int32") if _non_static_mode() else fill_constant( [0], dtype="int32", value="1") paddle.distributed.all_reduce(tmp, use_calc_stream=True) paddle.distributed.wait(tmp) return gp def is_initialized(): """ Check whether the distributed environment has been initialized Returns (bool): `True` if distributed environment has been initialized, otherwise `False`. Examples: .. code-block:: python # required: distributed import paddle print(paddle.distributed.is_initialized()) # False paddle.distributed.init_parallel_env() print(paddle.distributed.is_initialized()) # True """ global _group_map_by_name return _default_group_name in _group_map_by_name def destroy_process_group(group=None): """ Destroy a given group for communication Args: group (ProcessGroup, optional): The group to be destroyed. All of process groups, including the default group, will be destroyed and the distributed environment will be deinitialized. Returns : None Examples: .. code-block:: python # required: distributed import paddle paddle.distributed.init_parallel_env() group = paddle.distributed.new_group([0, 1]) paddle.distributed.destroy_process_group(group) print(paddle.distributed.is_initialized()) # True paddle.distributed.destroy_process_group() print(paddle.distributed.is_initialized()) # False """ global _group_map global _group_map_by_name pg = _get_default_group() if group is None else group assert _group_map.get(pg.id, None) is not None, "Invalid group." if group is None: _group_map.clear() _group_map_by_name.clear() _group_map_backend.clear() else: del _group_map[pg.id] del _group_map_by_name[pg.name] del _group_map_backend[pg] def wait(tensor, group=None, use_calc_stream=True): """ wait to sync stream for group. Args: tensor (Tensor): The Tensor used before sync. group (Group): The Group instance to perform sync. use_calc_stream (bool): Wether to use calculation stream (True) or communication stream (False). Default to True. Returns: None. Examples: .. code-block:: python import paddle paddle.distributed.init_parallel_env() tindata = paddle.randn(shape=[2, 3]) paddle.distributed.all_reduce(tindata, use_calc_stream=True) paddle.distributed.wait(tindata) """ if group is not None and not group.is_member(): return ring_id = 0 if group is None else group.id if use_calc_stream: _sync_calc_stream(tensor) else: _sync_comm_stream(tensor, ring_id) def _sync_calc_stream(tensor): if _non_static_mode(): return _C_ops.c_sync_calc_stream(tensor, tensor) op_type = 'c_sync_calc_stream' helper = LayerHelper(op_type, **locals()) helper.append_op( type=op_type, inputs={'X': [tensor]}, outputs={'Out': [tensor]}, ) def _sync_comm_stream(tensor, ring_id=0): if _non_static_mode(): return _C_ops.c_sync_comm_stream([tensor], [tensor], 'ring_id', ring_id) op_type = 'c_sync_comm_stream' helper = LayerHelper(op_type, **locals()) helper.append_op( type=op_type, inputs={'X': [tensor]}, outputs={'Out': [tensor]}, attrs={'ring_id': ring_id}, ) def broadcast(tensor, src, group=None, use_calc_stream=True): """ Broadcast a tensor from the source to all others. As shown below, 4 GPUs each start 4 processes and GPU0 owns data 0. Through broadcast operator, the data 0 will be sent to all GPUs from GPU0. .. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/broadcast.png :width: 800 :alt: broadcast :align: center Args: tensor (Tensor): The Tensor to send if current rank is the source, or the tensor to receive otherwise. Its data type should be float16, float32, float64, int32 or int64. src (int): The source rank. group (Group): The group instance return by new_group or None for global default group. use_calc_stream (bool): Wether to use calculation stream (True) or communication stream (False). Default to True. Returns: None. Examples: .. code-block:: python # required: distributed import numpy as np import paddle from paddle.distributed import init_parallel_env paddle.set_device('gpu:%d'%paddle.distributed.ParallelEnv().dev_id) init_parallel_env() if paddle.distributed.ParallelEnv().local_rank == 0: np_data = np.array([[4, 5, 6], [4, 5, 6]]) else: np_data = np.array([[1, 2, 3], [1, 2, 3]]) data = paddle.to_tensor(np_data) paddle.distributed.broadcast(data, 1) out = data.numpy() # [[1, 2, 3], [1, 2, 3]] """ if group is not None and not group.is_member(): return if not isinstance(src, int): raise ValueError("src should be int.") if in_dygraph_mode(): group = _get_default_group() if group is None else group gsrc = group.get_group_rank(src) assert gsrc >= 0, ("src rank out of group, need global rank") task = group.process_group.broadcast(tensor, gsrc) if use_calc_stream: task.wait() return None else: return task ring_id = ring_id = 0 if group is None else group.id gsrc = src if group is None else group.get_group_rank(src) assert gsrc >= 0, ("src rank out of group, need global rank") if _non_static_mode(): return _C_ops.c_broadcast(tensor, tensor, 'root', gsrc, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id) op_type = 'c_broadcast' check_variable_and_dtype( tensor, 'tensor', ['float16', 'float32', 'float64', 'int32', 'int64'], 'broadcast') helper = LayerHelper(op_type, **locals()) helper.append_op(type=op_type, inputs={'X': [tensor]}, outputs={'Out': [tensor]}, attrs={ 'root': gsrc, 'use_calc_stream': use_calc_stream, 'ring_id': ring_id, }) def all_reduce(tensor, op=ReduceOp.SUM, group=None, use_calc_stream=True): """ Reduce a tensor over all ranks so that all get the result. As shown below, 4 GPUs each start 4 processes and the data on each GPU is represnted by the GPU number. The reduce operator is sum. Through all_reduce operator, each GPU will have the sum of the data from all GPUs. .. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/allreduce.png :width: 800 :alt: all_reduce :align: center Args: tensor (Tensor): The input Tensor. It also works as the output Tensor. Its data type should be float16, float32, float64, int32 or int64. op (ReduceOp.SUM|ReduceOp.MAX|ReduceOp.Min|ReduceOp.PROD): Optional. The operation used. Default value is ReduceOp.SUM. group (Group): The group instance return by new_group or None for global default group. use_calc_stream (bool): Wether to use calculation stream (True) or communication stream (False). Default to True. Returns: None. Examples: .. code-block:: python # required: distributed import numpy as np import paddle from paddle.distributed import ReduceOp from paddle.distributed import init_parallel_env paddle.set_device('gpu:%d'%paddle.distributed.ParallelEnv().dev_id) init_parallel_env() if paddle.distributed.ParallelEnv().local_rank == 0: np_data = np.array([[4, 5, 6], [4, 5, 6]]) else: np_data = np.array([[1, 2, 3], [1, 2, 3]]) data = paddle.to_tensor(np_data) paddle.distributed.all_reduce(data) out = data.numpy() # [[5, 7, 9], [5, 7, 9]] """ if group is not None and not group.is_member(): return if in_dygraph_mode(): op_type = _get_reduce_op(op, "all_reduce") group = _get_default_group() if group is None else group task = group.process_group.allreduce(tensor, op_type) if use_calc_stream: task.wait() return None else: return task ring_id = 0 if group is None else group.id if _non_static_mode(): if op == ReduceOp.SUM: return _C_ops.c_allreduce_sum_(tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id) elif op == ReduceOp.MAX: return _C_ops.c_allreduce_max_(tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id) elif op == ReduceOp.MIN: return _C_ops.c_allreduce_min_(tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id) elif op == ReduceOp.PROD: return _C_ops.c_allreduce_prod_(tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id) else: raise ValueError("Unknown parameter: {}.".format(op)) check_variable_and_dtype( tensor, 'tensor', ['float16', 'float32', 'float64', 'int32', 'int64'], 'all_reduce') if op == ReduceOp.SUM: op_type = 'c_allreduce_sum' elif op == ReduceOp.MAX: op_type = 'c_allreduce_max' elif op == ReduceOp.MIN: op_type = 'c_allreduce_min' elif op == ReduceOp.PROD: op_type = 'c_allreduce_prod' if not isinstance(ring_id, int): raise ValueError("The type of 'ring_id' for all_reduce should be int.") helper = LayerHelper(op_type, **locals()) helper.append_op(type=op_type, inputs={'X': [tensor]}, outputs={'Out': [tensor]}, attrs={ 'ring_id': ring_id, 'use_calc_stream': use_calc_stream }) def reduce(tensor, dst, op=ReduceOp.SUM, group=None, use_calc_stream=True): """ Reduce a tensor to the destination from all others. As shown below, 4 GPUs each start 4 processes and the data on each GPU is respresnted by the GPU number. The destination of the reduce operator is GPU0 and the process is sum. Through reduce operator, the GPU0 will owns the sum of all data from all GPUs. .. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/reduce.png :width: 800 :alt: reduce :align: center Args: tensor (Tensor): The output Tensor for the destination and the input Tensor otherwise. Its data type should be float16, float32, float64, int32 or int64. dst (int): The destination rank id. op (ReduceOp.SUM|ReduceOp.MAX|ReduceOp.Min|ReduceOp.PROD): Optional. The operation used. Default value is ReduceOp.SUM. group (Group): The group instance return by new_group or None for global default group. use_calc_stream (bool): Wether to use calculation stream (True) or communication stream (False). Default to True. Returns: None. Examples: .. code-block:: python # required: distributed import numpy as np import paddle from paddle.distributed import init_parallel_env paddle.set_device('gpu:%d'%paddle.distributed.ParallelEnv().dev_id) init_parallel_env() if paddle.distributed.ParallelEnv().local_rank == 0: np_data = np.array([[4, 5, 6], [4, 5, 6]]) else: np_data = np.array([[1, 2, 3], [1, 2, 3]]) data = paddle.to_tensor(np_data) paddle.distributed.reduce(data, 0) out = data.numpy() # [[5, 7, 9], [5, 7, 9]] """ if group is not None and not group.is_member(): return if in_dygraph_mode(): op_type = _get_reduce_op(op, "reduce") group = _get_default_group() if group is None else group gdst = group.get_group_rank(dst) assert gdst >= 0, ("dst rank out of group, need global rank") task = group.process_group.reduce(tensor, gdst, op_type) if use_calc_stream: task.wait() return None else: return task ring_id = 0 if group is None else group.id gdst = dst if group is None else group.get_group_rank(dst) assert gdst >= 0, ("dst rank out of group, need global rank") if _non_static_mode(): if op == ReduceOp.SUM: return _C_ops.c_reduce_sum(tensor, tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id, 'root_id', gdst) elif op == ReduceOp.MAX: return _C_ops.c_reduce_max(tensor, tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id, 'root_id', gdst) elif op == ReduceOp.MIN: return _C_ops.c_reduce_min(tensor, tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id, 'root_id', gdst) elif op == ReduceOp.PROD: return _C_ops.c_reduce_prod(tensor, tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id, 'root_id', gdst) else: raise ValueError("Unknown parameter: {}.".format(op)) op_type = 'c_reduce' check_variable_and_dtype( tensor, 'tensor', ['float16', 'float32', 'float64', 'int32', 'int64'], 'all_reduce') if op == ReduceOp.SUM: op_type = 'c_reduce_sum' elif op == ReduceOp.MAX: op_type = 'c_reduce_max' elif op == ReduceOp.MIN: op_type = 'c_reduce_min' elif op == ReduceOp.PROD: op_type = 'c_reduce_prod' helper = LayerHelper(op_type, **locals()) helper.append_op(type=op_type, inputs={'X': [tensor]}, outputs={'Out': [tensor]}, attrs={ 'ring_id': ring_id, 'use_calc_stream': use_calc_stream, 'root_id': gdst, }) def all_gather(tensor_list, tensor, group=None, use_calc_stream=True): """ Gather tensors from all participators and all get the result. As shown below, 4 GPUs each start 4 processes and the data on each GPU is represnted by the GPU number. Through the all_gather operator, each GPU will have data from all GPUs. .. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/allgather.png :width: 800 :alt: all_gather :align: center Args: tensor_list (list): A list of output Tensors. Every element in the list must be a Tensor whose data type should be float16, float32, float64, int32, int64, int8, uint8, bool, complex64 or complex128. tensor (Tensor): The Tensor to send. Its data type should be float16, float32, float64, int32, int64, int8, uint8, bool, complex64 or complex128. group (Group): The group instance return by new_group or None for global default group. use_calc_stream (bool): Wether to use calculation stream (True) or communication stream (False). Default to True. Returns: None. Examples: .. code-block:: python # required: distributed import paddle from paddle.distributed import init_parallel_env paddle.set_device('gpu:%d'%paddle.distributed.ParallelEnv().dev_id) init_parallel_env() tensor_list = [] if paddle.distributed.ParallelEnv().local_rank == 0: data1 = paddle.to_tensor([[4, 5, 6], [4, 5, 6]]) paddle.distributed.all_gather(tensor_list, data1) else: data2 = paddle.to_tensor([[1, 2, 3], [1, 2, 3]]) paddle.distributed.all_gather(tensor_list, data2) """ if group is not None and not group.is_member(): return def convert_to_complex(list_of_tensor): list_of_complex = [] for tensor in list_of_tensor: list_of_complex.append(paddle.as_complex(tensor)) return list_of_complex is_input_complex = (tensor.dtype == paddle.complex64 or tensor.dtype == paddle.complex128) if is_input_complex: tensor = paddle.as_real(tensor) if in_dygraph_mode(): group = _get_default_group() if group is None else group if len(tensor_list) == 0: tensor_shape = list(tensor.shape) tensor_shape[0] *= group.nranks out = paddle.empty(tensor_shape, tensor.dtype) else: out = paddle.concat(tensor_list, axis=0) task = group.process_group.all_gather(tensor, out) task.wait() tensor_list.clear() list_of_tensor = paddle.split(out, group.nranks, 0) if is_input_complex: tensor_list.extend(convert_to_complex(list_of_tensor)) else: tensor_list.extend(list_of_tensor) return ring_id = 0 if group is None else group.id nranks = _get_global_group().nranks if group is None else group.nranks if _non_static_mode(): out = _C_ops.c_allgather(tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id, 'nranks', nranks) else: op_type = 'c_allgather' helper = LayerHelper(op_type, **locals()) out = helper.create_variable_for_type_inference(dtype=tensor.dtype) if not isinstance(tensor_list, list): raise ValueError("The type of 'tensor_list' for all_gather " "should be list.") for elem in tensor_list: check_variable_and_dtype(elem, 'tensor_list', [ 'float16', 'float32', 'float64', 'int32', 'int64', 'bool', 'int8', 'uint8', 'complex64', 'complex128' ], 'all_gather') check_variable_and_dtype(tensor, 'tensor', [ 'float16', 'float32', 'float64', 'int32', 'int64', 'bool', 'int8', 'uint8', 'complex64', 'complex128' ], 'all_gather') helper.append_op(type=op_type, inputs={'X': [tensor]}, outputs={'Out': [out]}, attrs={ 'ring_id': ring_id, 'use_calc_stream': use_calc_stream, 'nranks': nranks }) list_of_tensor = paddle.split(out, nranks, 0) if is_input_complex: tensor_list.extend(convert_to_complex(list_of_tensor)) else: tensor_list.extend(list_of_tensor) def _convert_object_to_tensor(obj): _pickler = pickle.Pickler f = io.BytesIO() _pickler(f).dump(obj) data = np.frombuffer(f.getvalue(), dtype=np.uint8) tensor = paddle.to_tensor(data) return tensor, tensor.numel() def _convert_tensor_to_object(tensor, len_of_tensor): _unpickler = pickle.Unpickler return _unpickler(io.BytesIO(tensor.numpy()[:len_of_tensor])).load() def all_gather_object(object_list, obj, group=None): """ Gather picklable objects from all participators and all get the result. Similiar to all_gather(), but python object can be passed in. Args: object_list (list): A list of output object. The datatype of every element in the list is same as the input obj. obj (Any): The picklable object to send. group (Group): The group instance return by new_group or None for global default group. Returns: None. Warning: This API only supports the dygraph mode. Examples: .. code-block:: python # required: distributed import paddle import paddle.distributed as dist paddle.set_device('gpu:%d'%paddle.distributed.ParallelEnv().dev_id) dist.init_parallel_env() object_list = [] if paddle.distributed.ParallelEnv().local_rank == 0: obj = {"foo": [1, 2, 3]} paddle.distributed.all_gather_object(object_list, obj) else: obj = {"bar": [4, 5, 6]} paddle.distributed.all_gather_object(object_list, obj) """ assert in_dygraph_mode( ), "all_gather_object doesn't support static graph mode." tensor, len_of_tensor = _convert_object_to_tensor(obj) # gather len_of_tensor from all ranks list_len_of_tensor = [] all_gather(list_len_of_tensor, len_of_tensor, group) # get the max length from list max_len_of_tensor = int(max(list_len_of_tensor).item()) # resize the input tensor to max length avoid hang in all gather # Note(liyurui): Maybe we should support various length all_gather? # Now this operation is efficient for we don't support resize in python. numpy_data = tensor.numpy() numpy_data = np.resize(numpy_data, [max_len_of_tensor]) input_tensor = paddle.to_tensor(numpy_data) tensor_list = [] all_gather(tensor_list, input_tensor, group) for i, tensor in enumerate(tensor_list): object_list.append( _convert_tensor_to_object(tensor, list_len_of_tensor[i])) def scatter(tensor, tensor_list=None, src=0, group=None, use_calc_stream=True): """ Scatter a tensor to all participators. As shown below, 4 GPUs each start 4 processes and the source of the scatter is GPU0. Through scatter operator, the data in GPU0 will be sent to all GPUs averagely. .. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/scatter.png :width: 800 :alt: scatter :align: center Args: tensor (Tensor): The output Tensor. Its data type should be float16, float32, float64, int32 or int64. tensor_list (list|tuple): A list/tuple of Tensors to scatter. Every element in the list must be a Tensor whose data type should be float16, float32, float64, int32 or int64. Default value is None. src (int): The source rank id. Default value is 0. group (Group): The group instance return by new_group or None for global default group. use_calc_stream (bool): Wether to use calculation stream (True) or communication stream (False). Default to True. Returns: None. Examples: .. code-block:: python # required: distributed import numpy as np import paddle from paddle.distributed import init_parallel_env paddle.set_device('gpu:%d'%paddle.distributed.ParallelEnv().dev_id) init_parallel_env() if paddle.distributed.ParallelEnv().local_rank == 0: np_data1 = np.array([7, 8, 9]) np_data2 = np.array([10, 11, 12]) else: np_data1 = np.array([1, 2, 3]) np_data2 = np.array([4, 5, 6]) data1 = paddle.to_tensor(np_data1) data2 = paddle.to_tensor(np_data2) if paddle.distributed.ParallelEnv().local_rank == 0: paddle.distributed.scatter(data1, src=1) else: paddle.distributed.scatter(data1, tensor_list=[data1, data2], src=1) out = data1.numpy() """ if group is not None and not group.is_member(): return if not isinstance(src, int): raise ValueError("src should be int.") if in_dygraph_mode(): group = _get_default_group() if group is None else group gsrc = group.get_group_rank(src) rank = group.rank nranks = group.nranks else: ring_id = 0 if group is None else group.id gsrc = src if group is None else group.get_group_rank(src) rank = _get_global_group().rank if group is None else group.rank nranks = _get_global_group().nranks if group is None else group.nranks assert gsrc >= 0, ("src rank out of group, need global rank") if rank != gsrc: tensor_list = [] for _ in range(nranks): tensor_list.append(tensor) temp = paddle.concat(tensor_list, axis=0) if in_dygraph_mode(): task = group.process_group.scatter(temp, tensor, gsrc) if use_calc_stream: task.wait() return None else: return task if _non_static_mode(): return _C_ops.c_scatter(temp, tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id, 'nranks', nranks, 'root', gsrc) op_type = 'c_scatter' check_variable_and_dtype( tensor, 'tensor', ['float16', 'float32', 'float64', 'int32', 'int64'], 'scatter') helper = LayerHelper(op_type, **locals()) helper.append_op(type=op_type, inputs={'X': [temp]}, outputs={'Out': [tensor]}, attrs={ 'ring_id': ring_id, 'root': gsrc, 'use_calc_stream': use_calc_stream, 'nranks': nranks, }) def _c_identity(tensor, group=None): """ Return a copy of the tensor, mainly used with model parallel. Args: tensor (Tensor): The input Tensor. Its data type should be float16, float32, float64, int32 or int64. group (int): The id of the process group to work on. Returns: Tensor. """ if group is not None and not group.is_member(): return ring_id = 0 if group is None else group.id if _non_static_mode(): return _C_ops.c_identity(tensor, 'use_calc_stream', True, 'ring_id', ring_id, 'use_model_parallel', True) op_type = 'c_identity' helper = LayerHelper(op_type, **locals()) out = helper.create_variable_for_type_inference(dtype=tensor.dtype) check_variable_and_dtype( tensor, 'tensor', ['float16', 'float32', 'float64', 'int32', 'int64'], '_c_identity') helper.append_op(type=op_type, inputs={'X': tensor}, outputs={'Out': out}, attrs={ 'ring_id': ring_id, 'use_calc_stream': True, 'use_model_parallel': True, }) return out def _c_concat(tensor, group=None): """ Return allgather of the tensor, mainly used with model parallel. Args: tensor (Tensor): The input Tensor. Its data type should be float16, float32, float64, int32 or int64. group (int): The id of the process group to work on. Returns: Tensor. """ if group is not None and not group.is_member(): return group = _get_default_group() if group is None else group ring_id = group.id global_rank = _get_global_env().rank rank = group.rank nranks = group.nranks if _non_static_mode(): return _C_ops.c_concat(tensor, 'ring_id', ring_id, 'use_calc_stream', True, 'rank', rank, 'nranks', nranks, 'use_model_parallel', True) op_type = 'c_concat' helper = LayerHelper(op_type, **locals()) out = helper.create_variable_for_type_inference(dtype=tensor.dtype) check_variable_and_dtype( tensor, 'tensor', ['float16', 'float32', 'float64', 'int32', 'int64'], '_c_concat') helper.append_op(type=op_type, inputs={'X': tensor}, outputs={'Out': out}, attrs={ 'ring_id': ring_id, 'use_calc_stream': True, 'use_model_parallel': True, 'nranks': nranks, 'rank': rank }) return out def _c_split(tensor, group=None): """ Split tensor evenly among all members, mainly used with model parallel. Args: tensor (Tensor): The input Tensor. Its data type should be float16, float32, float64, int32 or int64. rank (int): The rank of the current process. group (int): The id of the process group to work on. Returns: Tensor. """ if group is not None and not group.is_member(): return ring_id = 0 if group is None else group.id global_rank = _get_global_env().rank rank = global_rank if group is None else group.get_group_rank(global_rank) nranks = _get_global_env().world_size if group is None else group.nranks if _non_static_mode(): return _C_ops.c_split(tensor, 'use_calc_stream', True, 'ring_id', ring_id, 'rank', rank, 'nranks', nranks, 'use_model_parallel', True) op_type = 'c_split' helper = LayerHelper(op_type, **locals()) out = helper.create_variable_for_type_inference(dtype=tensor.dtype) check_variable_and_dtype( tensor, 'tensor', ['float16', 'float32', 'float64', 'int32', 'int64'], '_c_split') helper.append_op(type=op_type, inputs={'X': tensor}, outputs={'Out': out}, attrs={ 'ring_id': ring_id, 'use_calc_stream': True, 'rank': rank, 'nranks': nranks, 'use_model_parallel': True, }) return out def _mp_allreduce(tensor, op=ReduceOp.SUM, group=None, use_calc_stream=True, use_model_parallel=True): """[it is same as allreduce above, but it supports model parallel. And it support inplace startegy] """ if group is not None and not group.is_member(): return ring_id = 0 if group is None else group.id if in_dygraph_mode(): assert op == ReduceOp.SUM, "Unknown parameter: {}.".format(op) from paddle.autograd import PyLayer class mp_allreduce_eager(PyLayer): @staticmethod def forward(ctx, tensor, use_calc_stream, ring_id, use_model_parallel): ctx.ring_id = ring_id return _C_ops.c_allreduce_sum_(tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id, "use_model_parallel", use_model_parallel) @staticmethod def backward(ctx, dy): return _C_ops.c_identity(dy, 'use_calc_stream', True, 'ring_id', ctx.ring_id, 'use_model_parallel', True) return mp_allreduce_eager.apply(tensor, use_calc_stream, ring_id, use_model_parallel) elif _in_legacy_dygraph(): if op == ReduceOp.SUM: return _C_ops.c_allreduce_sum_(tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id, "use_model_parallel", use_model_parallel) else: raise ValueError("Unknown parameter: {}.".format(op)) op_type = 'c_allreduce_sum' helper = LayerHelper(op_type, **locals()) out = helper.create_variable_for_type_inference(dtype=tensor.dtype) check_variable_and_dtype( tensor, 'tensor', ['float16', 'float32', 'float64', 'int32', 'int64'], op_type) helper.append_op(type=op_type, inputs={'X': tensor}, outputs={'Out': out}, attrs={ 'ring_id': ring_id, 'use_calc_stream': use_calc_stream, 'use_model_parallel': use_model_parallel, }) return out def _c_lookup_table(table, index, start_index=0, name=None): """ Lookup table according to index. Args: table (Tensor): The input Tensor. Its data type should be float16, float32, float64. index (Tensor): The index to lookup table. start_index (int): The initial index for table range. name (string): The name of the api Returns: Tensor. """ if _non_static_mode(): return _C_ops.c_embedding(table, index, "start_index", start_index) op_type = 'c_embedding' helper = LayerHelper(op_type, **locals()) dtype = helper.input_dtype(input_param_name='table') check_variable_and_dtype(index, 'input', ['int32', 'int64'], op_type) tmp = helper.create_variable_for_type_inference(dtype) helper.append_op(type='c_embedding', inputs={ 'Ids': index, 'W': table }, outputs={'Out': tmp}, attrs={"start_index": start_index}) return tmp class _Linear(layers.Layer): """ Linear """ def __init__(self, in_features, out_features, weight_attr=None, bias_attr=None, name=None): super(_Linear, self).__init__() self._dtype = self._helper.get_default_dtype() self._weight_attr = weight_attr self._bias_attr = bias_attr self.weight = self.create_parameter(shape=[in_features, out_features], attr=self._weight_attr, dtype=self._dtype, is_bias=False) self.bias = self.create_parameter(shape=[out_features], attr=self._bias_attr, dtype=self._dtype, is_bias=True) self.name = name def forward(self, input): out = _linear(x=input, weight=self.weight, bias=self.bias, name=self.name) return out def extra_repr(self): name_str = ', name={}'.format(self.name) if self.name else '' return 'in_features={}, out_features={}, dtype={}{}'.format( self.weight.shape[0], self.weight.shape[1], self._dtype, name_str) def _c_softmax_with_cross_entropy(logits, label, group=None, return_softmax=False): if group is not None and not group.is_member(): return ring_id = 0 if group is None else group.id global_rank = _get_global_env().rank rank = global_rank if group is None else group.get_group_rank(global_rank) nranks = _get_global_env().world_size if group is None else group.nranks input_dims = len(list(logits.shape)) label_dims = len(list(label.shape)) if input_dims - 1 != label_dims and input_dims != label_dims: raise ValueError( 'Expected nput_dims - 1 = label_dims or input_dims == label_dims\ (got nput_dims{}, label_dims{})'.format(input_dims, label_dims)) if input_dims - 1 == label_dims: label = paddle.unsqueeze(label, axis=-1) if _non_static_mode(): softmax, loss = _C_ops.c_softmax_with_cross_entropy( logits, label, 'ring_id', ring_id, 'rank', rank, 'nranks', nranks) if not return_softmax: return loss else: return loss, softmax attrs = { 'ring_id': ring_id, 'rank': rank, 'nranks': nranks, } helper = LayerHelper('c_softmax_with_cross_entropy', **locals()) softmax = helper.create_variable_for_type_inference(dtype=logits.dtype) loss = helper.create_variable_for_type_inference(dtype=logits.dtype) helper.append_op(type='c_softmax_with_cross_entropy', inputs={ 'Logits': logits, 'Label': label }, outputs={ 'Softmax': softmax, 'Loss': loss }, attrs=attrs) if return_softmax: return loss, softmax return loss def _linear(x, weight, bias=None, name=None): """ Fuction Linear """ if _non_static_mode(): pre_bias = _varbase_creator(dtype=x.dtype) _C_ops.matmul(x, weight, pre_bias, 'transpose_X', False, 'transpose_Y', False, "alpha", 1) return dygraph_utils._append_bias_in_dygraph(pre_bias, bias, axis=len(x.shape) - 1) else: helper = LayerHelper('linear', **locals()) dtype = x.dtype assert len( x.shape) < 4, "X latitude is not supported greater than 3 now." check_variable_and_dtype(x, 'x', ['float16', 'float32', 'float64'], 'linear') check_dtype(dtype, 'dtype', ['float16', 'float32', 'float64'], 'linear') inputs = {'X': [x], 'Y': [weight]} attrs = { 'transpose_X': False, 'transpose_Y': False, 'alpha': 1, } tmp = helper.create_variable_for_type_inference(dtype) helper.append_op(type='matmul_v2', inputs=inputs, outputs={'Out': tmp}, attrs=attrs) if bias is not None: res = helper.create_variable_for_type_inference(dtype) helper.append_op(type='elementwise_add', inputs={ 'X': [tmp], 'Y': [bias] }, outputs={'Out': [res]}, attrs={'axis': len(x.shape) - 1}) else: res = tmp return res def _set_var_distributed(var): if var is None: return var.is_distributed = True # NOTE: use current_block and find_var_recursive to support while_loop startup_block = paddle.static.default_startup_program().current_block() main_block = paddle.static.default_main_program().current_block() startup_block._find_var_recursive(var.name).is_distributed = True main_block._find_var_recursive(var.name).is_distributed = True def _parallel_linear(x, num_rows, num_cols, axis, param_attr, bias_attr, gather_out, inner_rank, nranks, split_tensor, name, group=None): """ Parallel Linear axis the dimension of the parameter of linear layer. axis = 0: the row dimension axis = 1: the col dimension """ if group is not None and not group.is_member(): return ring_id = 0 if group is None else group.id if axis == 0: if split_tensor: x = _c_split(x, group=group) else: x = _c_identity(x, group=group) linear = paddle.nn.Linear(num_rows, num_cols, weight_attr=param_attr, bias_attr=bias_attr, name=name) # NOTE: npu linear function use matmul_v2 but linear use matmul linear_function = _linear if core.is_compiled_with_npu()\ else paddle.nn.functional.linear linear_out = linear_function( x, linear.weight, # NOTE(wangxi): row split, bias need add after allreduce None if axis == 0 else linear.bias, linear.name) _set_var_distributed(linear.weight) # set is_distributed for splited bias # if a linear layer is splited by row, each rank would hold a complete bias and they should be the same in each rank. # if a linear layer is splited by col, the bias would also be split into each rank as its weight if axis == 1 and linear._bias_attr != False: _set_var_distributed(linear.bias) if not gather_out: return linear_out out_shape = list(linear_out.shape) out_shape[0] *= 1 if axis == 0 else nranks main_block = paddle.static.default_main_program().current_block() out = main_block.create_var( shape=out_shape, dtype=linear_out.dtype, type=linear_out.type, lod_level=linear_out.lod_level, persistable=False, is_data=False, need_check_feed=linear_out.desc.need_check_feed()) if axis == 0: main_block.append_op(type='c_allreduce_sum', inputs={'X': linear_out}, outputs={'Out': out}, attrs={ 'ring_id': ring_id, 'use_calc_stream': True, 'use_model_parallel': True }) if linear.bias is not None: out = out + linear.bias else: main_block.append_op(type='c_concat', inputs={'X': linear_out}, outputs={'Out': out}, attrs={ 'rank': inner_rank, 'ring_id': ring_id, 'nranks': nranks, 'use_calc_stream': True, 'use_model_parallel': True }) return out def _parallel_embedding(x, per_part_embeddings, origin_size, param_attr, inner_rank, num_partitions, name, group=None): """ Parallel Embedding """ if group is not None and not group.is_member(): return ring_id = 0 if group is None else group.id helper = LayerHelper("_parallel_embedding", **locals()) per_part_size = per_part_embeddings rank = inner_rank vocab_start_index = rank * per_part_size dtype = helper.get_default_dtype() size = [per_part_size, origin_size[1]] weight = helper.create_parameter(attr=param_attr, shape=size, dtype=dtype, is_bias=False) if num_partitions == 1: return paddle.nn.functional.embedding(x, weight=weight, padding_idx=None, sparse=False, name=name) startup_block = paddle.static.default_startup_program().global_block() main_block = paddle.static.default_main_program().global_block() startup_block.vars[weight.name].is_distributed = True main_block.vars[weight.name].is_distributed = True output_parallel = paddle.distributed.collective._c_lookup_table( weight, x, start_index=vocab_start_index, name=name) out = paddle.distributed.collective._mp_allreduce(output_parallel, group=group, use_calc_stream=True, use_model_parallel=True) return out def split(x, size, operation, axis=0, num_partitions=1, gather_out=True, weight_attr=None, bias_attr=None, name=None): """ Split the weight of the specified operation into multiple devices and do the computation in parallel. Now the following three cases are supported. Case 1: Parallel Embedding The weight of the embedding operation is a NxM matrix with N rows and M columns. With parallel embedding, the weight is split into num_partitions partitions, each of which is a matrix with (N/num_partitions + 1) rows and M column where the last row as the padding idx. Suppose we split the NxM weight into two partitons on device_0 and device_1 respectively. Then, one each device, the final weight has (N/2 + 1) rows with the index range from 0 to N/2. On device_0, all values in the input within [0, N/2 -1] keep unchanged and all other values are changed to N/2 which is the padding index and are mapped to all zeros after embedding. In the same way, on device_1, the value V in the input within [N/2, N-1] will be changed to (V - N/2), and all other values are changed to N/2 and are mapped to all zeros after embedding. Finally, the results on the two devices are sum-reduced. The Embedding put on single card is as shown below: .. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/split_embedding_single.png :width: 800 :height: 350 :alt: single_embedding :align: center Parallel Embedding is shown as below: .. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/split_embedding_split.png :width: 800 :alt: split_embedding :align: center Case 2: Row Parallel Linear The weight of the linear operation is a NxM matrix with N rows and M columns. With row parallel linear, the weight is split into num_partitions partitions, each of which is a matrix with N/num_partitions rows and M column. The linear layer put on single card is shown as below, the input variable is represented by X, the weight matrix is represented by W and the output vaiable is O. The linear layer on single card is simple matrix multiplication operation, O = X * W. .. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/split_single.png :width: 800 :alt: single_linear :align: center Row Parallel Linear is shown as below. As the name suggests, Row Parallel Linear splits the weight matrix W into [[W_row1], [W_row2]] along the row. And accordingly the input is splitted along the column into [X_col1, X_col2] and multiply their respective weight matrices. Finally apply AllReduce on the output from each card to get the final output. .. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/split_row.png :width: 800 :alt: split_row :align: center Case 3: Column Parallel Linear The weight of the linear operation is a NxM matrix with N rows and M columns. With column parallel linear, the weight is split into num_paratitions partitions, each of which is a matrix with N rows and M/num_partitions column. The linear layer put on single card has been illustrated on case 2 and Column Parallel Linear is shown as below. The Column Parallel Linear splits the weight matrix W into [W_col1, W_col2] along the column and these splitted matrices respectively multiply the input. Finally apply AllGather on the output from each card to get the final output. .. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/split_col.png :width: 800 :alt: split_col :align: center As observed, the column parallel linear and row parallel linear can be combined to skip one ALLGATHER communication operator. Furthermore the Attention and MLP can be combined to imporve the performance as shown below. .. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/split_col_row.png :width: 800 :alt: split_col_row :align: center Args: x (Tensor): Input tensor. It's data type should be float16, float32, float64, int32 or int64. size (list|tuple): A list or tuple with two elements indicating the shape of the weight. operation (str): The name of the operation. The supported operations are 'linear' and 'embedding'. axis (int, Optional): Indicate along which axis to split the weight. Default: 0. num_partitions (int, Optional): How many parts the weight is partitioned. Default: 1. gather_out (bool, Optional): Whether to gather the output after computation. By default, the output on each partitions will be gathered after computation. Default: True. weight_attr (ParamAttr, Optional): The parameter attribute for the learnable weights(Parameter) of the specified operation. Default: None. bias_attr (ParamAttr, Optional): The parameter attribute for the bias of the specified operation. Default: None. name (str, Optional): The default value is None. Normally there is no need for user to set this property. Default: None. For more information, please refer to :ref:`api_guide_Name`. Returns: Tensor. Examples: .. code-block:: python # required: distributed import paddle import paddle.distributed.fleet as fleet paddle.enable_static() paddle.set_device('gpu:%d'%paddle.distributed.ParallelEnv().dev_id) fleet.init(is_collective=True) data = paddle.randint(0, 8, shape=[10,4]) emb_out = paddle.distributed.split( data, (8, 8), operation="embedding", num_partitions=2) """ assert isinstance( size, (list, tuple)), ("The type of size for " "paddle.distributed.split must be list or tuple.") assert len(size) == 2, ("Number of elements in size of " "paddle.distributed.split must be two.") assert isinstance(operation, str), ("The type of operation for " "paddle.distributed.split must be str.") supported_operations = [ 'linear', 'embedding', ] assert operation in supported_operations, ( "The operation for " "paddle.distributed.split must be one of {}.".format( supported_operations)) if _non_static_mode(): raise ValueError( "paddle.distributed.split cannot be used in dynamic " "graph mode, plese use ParallelEmbedding, ParallelRowLinear, " "ParallelColumnLinear instead.") else: from .fleet import fleet assert fleet._role_maker, ("To use paddle.distributed.split, " "you must call fleet.init() firstly.") rank = fleet.worker_index() nranks = fleet.worker_num() # rank within a model parallel group inner_rank = rank % num_partitions if operation == "embedding": assert axis == 0, ("We only support to split the weight of embedding " "along the first axis now.") assert size[0] % num_partitions == 0, \ "The length of the vocabulary must be divisible by num_partitions " \ "but received vocabulary={} num_partitions={}".format(size[0], num_partitions) per_part_size = size[0] // num_partitions emb_out = _parallel_embedding(x, per_part_size, size, weight_attr, inner_rank, num_partitions, name, group=None) return emb_out else: should_split = False if axis == 0: assert size[0] % num_partitions == 0, ( "Number of rows of the weight for linear ({}) must be" " divisible by num_partitions ({})".format( size[0], num_partitions)) per_part_size = size[0] // num_partitions linear_size = (per_part_size, size[1]) if x.shape[-1] == size[0]: should_split = True elif axis == 1: assert size[1] % num_partitions == 0, ( "Number of column of the weight for linear ({}) must be" " divisible by num_partitions ({})".format( size[1], num_partitions)) per_part_size = size[1] // num_partitions linear_size = (size[0], per_part_size) else: raise ValueError("The value of axis must be 0 or 1, but the value " "given is {}.".format(axis)) linear_out = _parallel_linear(x, linear_size[0], linear_size[1], axis, weight_attr, bias_attr, gather_out, inner_rank, num_partitions, should_split, name=name, group=None) return linear_out def alltoall(in_tensor_list, out_tensor_list, group=None, use_calc_stream=True): """ Scatter tensors in in_tensor_list to all participators averagely and gather the result tensors in out_tensor_list. As shown below, the in_tensor_list in GPU0 includes 0_0 and 0_1, and GPU1 includes 1_0 and 1_1. Through alltoall operator, the 0_0 in GPU0 will be sent to GPU0 and 0_1 to GPU1, 1_0 in GPU1 sent to GPU0 and 1_1 to GPU1. Finally the out_tensor_list in GPU0 includes 0_0 and 1_0, and GPU1 includes 0_1 and 1_1. .. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/alltoall.png :width: 800 :alt: alltoall :align: center Args: in_tensor_list (list): A list of input Tensors. Every element in the list must be a Tensor whose data type should be float16, float32, float64, int32 or int64. out_tensor_list (list): A list of output Tensors. The data type of its elements should be the same as the data type of the input Tensors. group (Group, optional): The group instance return by new_group or None for global default group. Default: None. use_calc_stream (bool, optional): Whether to use calculation stream (True) or communication stream. Default: True. Returns: None. Examples: .. code-block:: python # required: distributed import numpy as np import paddle from paddle.distributed import init_parallel_env init_parallel_env() out_tensor_list = [] if paddle.distributed.ParallelEnv().rank == 0: np_data1 = np.array([[1, 2, 3], [4, 5, 6]]) np_data2 = np.array([[7, 8, 9], [10, 11, 12]]) else: np_data1 = np.array([[13, 14, 15], [16, 17, 18]]) np_data2 = np.array([[19, 20, 21], [22, 23, 24]]) data1 = paddle.to_tensor(np_data1) data2 = paddle.to_tensor(np_data2) paddle.distributed.alltoall([data1, data2], out_tensor_list) # out for rank 0: [[[1, 2, 3], [4, 5, 6]], [[13, 14, 15], [16, 17, 18]]] # out for rank 1: [[[7, 8, 9], [10, 11, 12]], [[19, 20, 21], [22, 23, 24]]] """ if group is not None and not group.is_member(): return if in_dygraph_mode(): group = _get_default_group() if group is None else group else: ring_id = 0 if group is None else group.id temp = paddle.concat(in_tensor_list, axis=0) nranks = len(in_tensor_list) if in_dygraph_mode(): if len(out_tensor_list) == 0: tensor_shape = list(in_tensor_list[0].shape) tensor_shape[0] *= nranks out = paddle.empty(tensor_shape, in_tensor_list[0].dtype) else: out = paddle.concat(out_tensor_list, axis=0) task = group.process_group.alltoall(temp, out) task.wait() out_tensor_list.clear() out_tensor_list.extend(paddle.split(out, nranks, 0)) return if _non_static_mode(): out = _C_ops.alltoall(temp, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id) else: op_type = 'alltoall' helper = LayerHelper(op_type, **locals()) out = helper.create_variable_for_type_inference( dtype=in_tensor_list[0].dtype) if not isinstance(in_tensor_list, list): raise ValueError("The type of 'in_tensor_list' for all_to_all " "should be list.") for elem in in_tensor_list: check_variable_and_dtype( elem, 'in_tensor_list', ['float16', 'float32', 'float64', 'int32', 'int64'], 'all_to_all') if not isinstance(out_tensor_list, list): raise ValueError("The type of 'out_tensor_list' for all_to_all " "should be list.") if len(out_tensor_list) != 0: raise ValueError("The 'out_tensor_list' for all_to_all " "must be an empty list.") helper.append_op(type=op_type, inputs={'X': [temp]}, outputs={'Out': [out]}, attrs={ 'ring_id': ring_id, 'use_calc_stream': use_calc_stream, }) out_tensor_list.extend(paddle.split(out, nranks, 0)) def alltoall_single(in_tensor, out_tensor, in_split_sizes=None, out_split_sizes=None, group=None, use_calc_stream=True): """ Scatter a single input tensor to all participators and gather the received tensors in out_tensor. .. note:: ``alltoall_single`` is only supported in eager mode. Args: in_tensor (Tensor): Input tensor. The data type should be float16, float32, float64, int32 or int64. out_tensor (Tensor): Output Tensor. The data type should be the same as the data type of the input Tensor. in_split_sizes (list[int], optional): Split sizes of ``in_tensor`` for dim[0]. If not given, dim[0] of ``in_tensor`` must be divisible by group size and ``in_tensor`` will be scattered averagely to all participators. Default: None. out_split_sizes (list[int], optional): Split sizes of ``out_tensor`` for dim[0]. If not given, dim[0] of ``out_tensor`` must be divisible by group size and ``out_tensor`` will be gathered averagely from all participators. Default: None. group (Group, optional): The group instance return by ``new_group`` or None for global default group. Default: None. use_calc_stream (bool, optional): Whether to use calculation stream (True) or communication stream. Default: True. Returns: None, if ``use_calc_stream`` is set to ``True``; ``Task`` of ``group``, if ``use_calc_stream`` is set to ``False``. Examples: .. code-block:: python # required: distributed import paddle import paddle.distributed as dist dist.init_parallel_env() rank = dist.get_rank() size = dist.get_world_size() # case 1 input = paddle.arange(2, dtype='int64') + rank * 2 # input for rank 0: [0, 1] # input for rank 1: [2, 3] output = paddle.empty([2], dtype='int64') dist.alltoall_single(input, output) # output for rank 0: [0, 2] # output for rank 1: [1, 3] # case 2 in_split_sizes = [i + 1 for i in range(size)] # in_split_sizes for rank 0: [1, 2] and for rank 1: [1, 2] out_split_sizes = [rank + 1 for i in range(size)] # out_split_sizes for rank 0: [1, 1] and for rank 1: [2, 2] input = paddle.ones([sum(in_split_sizes), size], dtype='float32') * rank # input for rank 0: [[0., 0.], [0., 0.], [0., 0.]] # input for rank 1: [[1., 1.], [1., 1.], [1., 1.]] output = paddle.empty([(rank + 1) * size, size], dtype='float32') group = dist.new_group([0, 1]) task = dist.alltoall_single(input, output, in_split_sizes, out_split_sizes, use_calc_stream=False, group=group) task.wait() # output for rank 0: [[0., 0.], [1., 1.]] # output for rank 1: [[0., 0.], [0., 0.], [1., 1.], [1., 1.]] """ if group is not None and not group.is_member(): return assert in_dygraph_mode(), "Only suppport alltoall_single in eager mode." # _check_single_tensor group = _get_default_group() if group is None else group in_split_sizes = [] if in_split_sizes is None else in_split_sizes out_split_sizes = [] if out_split_sizes is None else out_split_sizes task = group.process_group.alltoall_single(in_tensor, out_tensor, in_split_sizes, out_split_sizes) if use_calc_stream: task.wait() return else: return task def _get_group_rank(global_rank, group=None): return global_rank if group is None else group.get_group_rank(global_rank) def send(tensor, dst=0, group=None, use_calc_stream=True): """ Send a tensor to the receiver. Args: tensor (Tensor): The Tensor to send. Its data type should be float16, float32, float64, int32 or int64. dst (int): The destination rank id. group (Group, optional): The group instance return by new_group or None for global default group. Default: None. use_calc_stream (bool, optional): Whether to use calculate stream or communication stream. Default: True. Returns: None. Examples: .. code-block:: python # required: distributed import paddle from paddle.distributed import init_parallel_env init_parallel_env() if paddle.distributed.ParallelEnv().rank == 0: data = paddle.to_tensor([7, 8, 9]) paddle.distributed.send(data, dst=1) else: data = paddle.to_tensor([1,2,3]) paddle.distributed.recv(data, src=0) out = data.numpy() """ if group is not None and not group.is_member(): return dst = _get_group_rank(dst, group) if in_dygraph_mode(): group = _get_default_group() if group is None else group task = group.process_group.send(tensor, dst) if use_calc_stream: task.wait() return None else: return task ring_id = 0 if group is None else group.id if _non_static_mode(): return _C_ops.send_v2(tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id, 'peer', dst) op_type = 'send_v2' check_variable_and_dtype( tensor, 'tensor', ['float16', 'float32', 'float64', 'int32', 'int64'], 'send') helper = LayerHelper(op_type, **locals()) helper.append_op(type=op_type, inputs={'X': [tensor]}, attrs={ 'ring_id': ring_id, 'peer': dst, 'use_calc_stream': use_calc_stream, }) def recv(tensor, src=0, group=None, use_calc_stream=True): """ Receive a tensor to the sender. Args: tensor (Tensor): The Tensor to receive. Its data type should be float16, float32, float64, int32 or int64. src (int): The source rank id. group (Group, optional): The group instance return by new_group or None for global default group. Default: None. use_calc_stream (bool, optional): Whether to use calculate stream or communication stream. Default: True. Returns: None. Examples: .. code-block:: python # required: distributed import paddle from paddle.distributed import init_parallel_env init_parallel_env() if paddle.distributed.ParallelEnv().rank == 0: data = paddle.to_tensor([7, 8, 9]) paddle.distributed.send(data, dst=1) else: data = paddle.to_tensor([1,2,3]) paddle.distributed.recv(data, src=0) out = data.numpy() """ if group is not None and not group.is_member(): return src = _get_group_rank(src, group) if in_dygraph_mode(): group = _get_default_group() if group is None else group task = group.process_group.recv(tensor, src) if use_calc_stream: task.wait() return None else: return task ring_id = 0 if group is None else group.id if _non_static_mode(): return _C_ops.recv_v2(tensor, 'use_calc_stream', use_calc_stream, 'ring_id', ring_id, 'peer', src, 'dtype', tensor.dtype, 'out_shape', tensor.shape) op_type = 'recv_v2' check_variable_and_dtype( tensor, 'tensor', ['float16', 'float32', 'float64', 'int32', 'int64'], 'recv') helper = LayerHelper(op_type, **locals()) helper.append_op(type=op_type, outputs={'Out': [tensor]}, attrs={ 'ring_id': ring_id, 'peer': src, 'out_shape': tensor.shape, 'dtype': tensor.dtype, 'use_calc_stream': use_calc_stream, }) def _check_single_tensor(tensor, tensor_name): if not isinstance(tensor, (core.eager.Tensor, paddle.Tensor)): raise RuntimeError("Invalid function argument. Expected parameter {}" "to be of type paddle.Tensor, but it's {}".format( tensor_name, type(tensor))) def _check_tensor_list(tensor_list, tensor_name): if not isinstance(tensor_list, list) or \ not all(isinstance(t, (core.eager.Tensor, paddle.Tensor)) for t in tensor_list): raise RuntimeError("Invalid function argument. Expected parameter {}" "to be of type paddle.Tensor".format(tensor_name)) def isend(tensor, dst, group=None): """ Sends a tensor asynchronously Args: tensor (Tensor): The Tensor to send. Its data type should be float16, float32, float64, int32 or int64. dst (int): The destination rank. group (Group, optional): The group instance return by new_group or None for global default group. Default: None. Returns: A distributed task object. Warning: This API only supports the dygraph mode. Examples: .. code-block:: python # required: distributed import paddle import paddle.distributed as dist dist.init_parallel_env() rank = dist.get_rank() world_size = dist.get_world_size() if rank == 0: data = paddle.to_tensor([7, 8, 9]) task = paddle.distributed.isend(data, dst=1) else: data = paddle.to_tensor([1, 2, 3]) task = paddle.distributed.irecv(data, src=0) task.wait() print(data) # paddle.tensor([7, 8, 9]) # Rank-0 # paddle.tensor([7, 8, 9]) # Rank-1 """ _check_single_tensor(tensor, "tensor") if group is not None and not group.is_member(): return if in_dygraph_mode(): group = _get_default_group() if group is None else group group_dst_rank = group.get_group_rank(dst) assert group_dst_rank >= 0, ("dst rank out of group, need global rank") return group.process_group.send(tensor, group_dst_rank) else: raise RuntimeError("Don't support static graph mode currently.") def irecv(tensor, src=None, group=None): """ Receive a tensor to the sender. Args: tensor (Tensor): The Tensor to receive. Its data type should be float16, float32, float64, int32 or int64. src (int): The source rank id. group (Group, optional): The group instance return by new_group or None for global default group. Default: None. Returns: A distributed task object. Warning: This API only supports the dygraph mode. Examples: .. code-block:: python # required: distributed import paddle import paddle.distributed as dist dist.init_parallel_env() rank = dist.get_rank() world_size = dist.get_world_size() if rank == 0: data = paddle.to_tensor([7, 8, 9]) task = paddle.distributed.isend(data, dst=1) else: data = paddle.to_tensor([1, 2, 3]) task = paddle.distributed.irecv(data, src=0) task.wait() print(data) # paddle.tensor([7, 8, 9]) # Rank-0 # paddle.tensor([7, 8, 9]) # Rank-1 """ _check_single_tensor(tensor, "tensor") if group is not None and not group.is_member(): return if in_dygraph_mode(): group = _get_default_group() if group is None else group group_src_rank = group.get_group_rank(src) assert group_src_rank >= 0, ("src rank out of group, need global rank") return group.process_group.recv(tensor, group_src_rank) else: raise RuntimeError("Don't support static graph mode currently.") class P2POp(object): """ A class that makes point-to-point operations for "batch_isend_irecv". This class creates the type of P2P operation, communication buffer, peer rank, Group. Instances of this class will be passed to ``paddle.distributed.batch_isend_irecv`` for point-to-point communication. Args: op (callable): A function to send data to or receive data from a peer process. The type of ``op`` is either ``paddle.distributed.isend`` or ``paddle.distributed.irecv``. tensor (Tensor): Tensor to send or receive. peer (int): The destination or source rank. group (Group, optional): The group instance return by new_group or None for global default group. Default: None. """ def __init__(self, op, tensor, peer, group=None): if op not in [isend, irecv]: raise RuntimeError("Invalid ``op`` function. Expected ``op`` " "to be of type ``paddle.distributed.isend`` or " "``paddle.distributed.irecv``.") _check_single_tensor(tensor, "tensor") self.op = op self.tensor = tensor self.peer = peer self.group = _get_default_group() if group is None else group @contextlib.contextmanager def _with_batch_p2p_guard(backend): if backend == "nccl": core.ProcessGroupNCCL.group_start() try: yield finally: if backend == "nccl": core.ProcessGroupNCCL.group_end() def _check_p2p_op_list(p2p_op_list): """ Helper to check that the ``p2p_op_list`` is a list of P2POp instances and all ops use the same backend. """ if not isinstance(p2p_op_list, list) or not all( isinstance(p2p_op, P2POp) for p2p_op in p2p_op_list): raise RuntimeError("Invalid ``p2p_op_list``. Each op is expected to " "to be of type ``paddle.distributed.P2POp``.") backend = _group_map_backend[p2p_op_list[0].group] if not all(backend == _group_map_backend[p2p_op.group] for p2p_op in p2p_op_list): raise RuntimeError("All groups need to use the same backend.") def batch_isend_irecv(p2p_op_list): """ Send or Receive a batch of tensors asynchronously and return a list of requests. Process each of the point-to-point operations in ``p2p_op_list`` and return the corresponding tasks. NCCL are currently supported. Args: p2p_op_list: A list of point-to-point operations(type of each operator is ``paddle.distributed.P2POp``). The order of the isend/irecv in the list matters and it needs to match with corresponding isend/irecv on the remote end. Returns: A list of distributed tasks returned by calling the corresponding op in the op_list. Warning: This API only supports the dygraph mode. Examples: .. code-block:: python # required: distributed import paddle import paddle.distributed as dist dist.init_parallel_env() rank = dist.get_rank() world_size = dist.get_world_size() send_t = paddle.arange(2) + rank # paddle.tensor([0, 1]) # Rank-0 # paddle.tensor([1, 2]) # Rank-1 recv_t = paddle.empty(shape=[2], dtype=send_t.dtype) send_op = dist.P2POp(dist.isend, send_t, (rank + 1) % world_size) recv_op = dist.P2POp(dist.irecv, recv_t, (rank - 1 + world_size) % world_size) tasks = dist.batch_isend_irecv([send_op, recv_op]) for task in tasks: task.wait() print(recv_t) # paddle.tensor([1, 2]) # Rank-0 # paddle.tensor([0, 1]) # Rank-1 """ _check_p2p_op_list(p2p_op_list) group = p2p_op_list[0].group if group is not None and not group.is_member(): return if in_dygraph_mode(): group = _get_default_group() if group is None else group backend = _group_map_backend[group] tasks = [] with _with_batch_p2p_guard(backend): for p2p_op in p2p_op_list: op = p2p_op.op tensor = p2p_op.tensor peer = p2p_op.peer comm_group = p2p_op.group task = op(tensor, peer, comm_group) if task is not None: tasks.append(task) return tasks else: raise RuntimeError("Don't support static graph mode currently.") def reduce_scatter(tensor, tensor_list, op=ReduceOp.SUM, group=None, use_calc_stream=True): """ Reduces, then scatters a list of tensors to all processes in a group Args: tensor (Tensor): Output tensor. tensor_list (list[Tensor]): List of tensors to reduce and scatter. op (ReduceOp.SUM|ReduceOp.MAX|ReduceOp.Min|ReduceOp.PROD): Optional. The operation used. Default: ReduceOp.SUM. group (Group, optional): The group instance return by new_group or None for global default group. Default: None. use_calc_stream (bool, optional): Whether this op should be an async op. Returns: Async task handle, if use_calc_stream is set to False. None, if use_calc_stream or if not part of the group. Warning: This API only supports the dygraph mode. Examples: .. code-block:: python # required: distributed import paddle import paddle.distributed as dist dist.init_parallel_env() rank = dist.get_rank() world_size = dist.get_world_size() if rank == 0: t1 = paddle.to_tensor([0, 1]) t2 = paddle.to_tensor([2, 3]) else: t1 = paddle.to_tensor([4, 5]) t2 = paddle.to_tensor([6, 7]) tensor_list = [t1, t2] output = paddle.empty(shape=[2], dtype=tensor_list[0].dtype) dist.reduce_scatter(output, tensor_list) print(output) # [4, 6] # Rank-0 # [8, 10] # Rank-1 """ _check_single_tensor(tensor, "tensor") _check_tensor_list(tensor_list, "tensor_list") if group is not None and not group.is_member(): return if in_dygraph_mode(): op_type = _get_reduce_op(op, "reduce_scatter") group = _get_default_group() if group is None else group temp = paddle.concat(tensor_list, axis=0) task = group.process_group._reduce_scatter_base(tensor, temp, op_type) if use_calc_stream: task.wait() return None else: return task else: raise RuntimeError("Don't support static graph mode currently.") def _reduce_scatter_base(output, input, op=ReduceOp.SUM, group=None, use_calc_stream=True): """ Reduces, then scatters a flattened tensor to all processes in a group. Args: output (Tensor): Output tensor. input (Tensor): Input tensor that is of size output tensor size times world size op (ReduceOp.SUM|ReduceOp.MAX|ReduceOp.Min|ReduceOp.PROD): Optional. The operation used. Default: ReduceOp.SUM. group (ProcessGroup, optional): The process group to work on. If None, the default process group will be used. use_calc_stream (bool, optional): Wether to use calculation stream (True) or communication stream (False). Default to True. Returns: Async task handle, if use_calc_stream is set to False. None, if use_calc_stream or if not part of the group. Examples: .. code-block:: python # required: distributed import paddle import paddle.distributed as dist dist.init_parallel_env() rank = dist.get_rank() world_size = dist.get_world_size() input = paddle.arange(4) + rank # [0, 1, 2, 3] # Rank-0 # [1, 2, 3, 4] # Rank-1 output = paddle.empty(shape=[2], dtype=input.dtype) paddle.distributed.collective._reduce_scatter_base(output, input) print(output) # [1, 3] # Rank-0 # [5, 7] # Rank-1 """ _check_single_tensor(output, "output") _check_single_tensor(input, "input") if group is not None and not group.is_member(): return if in_dygraph_mode(): op_type = _get_reduce_op(op, "_reduce_scatter_base") group = _get_default_group() if group is None else group task = group.process_group._reduce_scatter_base(output, input, op_type) if use_calc_stream: task.wait() return None else: return task else: raise RuntimeError("Don't support static graph mode currently.")