parallel_executor.py 9.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
#   Copyright (c) 2018 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 core
import multiprocessing
import framework
import executor
J
JiayiFeng 已提交
19
import warnings
Y
Yu Yang 已提交
20
import sys
C
chengduoZH 已提交
21
import os
22

Y
yuyang18 已提交
23
__all__ = ['ParallelExecutor', 'ExecutionStrategy', 'BuildStrategy']
Y
yuyang18 已提交
24 25

ExecutionStrategy = core.ParallelExecutor.ExecutionStrategy
Y
yuyang18 已提交
26
BuildStrategy = core.ParallelExecutor.BuildStrategy
27 28 29


class ParallelExecutor(object):
X
Xin Pan 已提交
30 31
    def __init__(self,
                 use_cuda,
32 33
                 loss_name=None,
                 main_program=None,
Y
Yu Yang 已提交
34
                 share_vars_from=None,
Y
yuyang18 已提交
35
                 exec_strategy=None,
Y
yuyang18 已提交
36
                 build_strategy=None,
T
typhoonzero 已提交
37
                 num_trainers=1,
38
                 trainer_id=0,
Y
yuyang18 已提交
39
                 **kwargs):
40 41 42 43 44 45 46 47 48 49
        """
        ParallelExecutor can run program in parallel.

        Args:
            use_cuda(bool): Whether to use CUDA or not.
            loss_name(str, default None): The loss name must set in training.
            main_program(Program, default None): The program that need to run,
                if not provided, then default_main_program will be used.
            share_vars_from(ParallelExecutor, default None): If provied,
                it will share variables from the specified ParallelExecutor.
T
typhoonzero 已提交
50
            num_trainers(int, default 1): If greater than 1, NCCL will be
T
typhoonzero 已提交
51 52
                initialized with multpile rank of nodes, each node should have
                same number of GPUs. Distributed training will be enabled then.
T
typhoonzero 已提交
53
            trainer_id(int, default 0): Must use together with num_trainers.
T
typhoonzero 已提交
54
                trainer_id is the "rank" of current node starts from 0.
55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72

        Returns:
            A ParallelExecutor object.

        Raises:
            TypeError: If share_vars_from is provided, but not ParallelExecutor
                object.

        Examples:
            .. code-block:: python

              train_exe = fluid.ParallelExecutor(
                  use_cuda=True, loss_name=loss.name)
              test_exe = fluid.ParallelExecutor(
                  use_cuda=True,
                  main_program=test_program,
                  share_vars_from=train_exe)

73 74
              train_loss, = train_exe.run([loss.name], feed=feed_dict)
              test_loss, = test_exe.run([loss.name], feed=feed_dict)
75
        """
Y
yuyang18 已提交
76 77 78 79 80 81 82 83
        if len(kwargs) != 0:
            err_msg = ""
            for key in kwargs:
                if key in dir(ExecutionStrategy):
                    err_msg += \
                        "Setting {0} by constructor is deprecated. Use " \
                        "strategy=ExecutionStrategy(); strategy.{0}=xxx; " \
                        "pe=ParallelExecutor(exec_strategy=strategy) " \
Y
yuyang18 已提交
84 85 86 87 88 89 90 91 92 93
                        "instead.\n ".format(key)
                elif key in dir(BuildStrategy):
                    err_msg += \
                        "Setting {0} by constructor is deprecated. Use " \
                        "strategy=BuildStrategy(); See help(" \
                        "paddle.fluid.ParallelExecutor.BuildStrategy) \n".format(
                            key)
                else:
                    err_msg += "Setting {0} by constructor is deprecated. Use strategy.\n".format(
                        key)
Y
yuyang18 已提交
94
            raise ValueError(err_msg)
95

X
Xin Pan 已提交
96 97
        self._places = []
        self._act_places = []
98 99 100
        if use_cuda:
            for i in xrange(core.get_cuda_device_count()):
                p = core.Place()
X
Xin Pan 已提交
101 102 103
                self._act_places.append(core.CUDAPlace(i))
                p.set_place(self._act_places[-1])
                self._places.append(p)
104
        else:
C
chengduoZH 已提交
105 106 107
            cpu_num = int(
                os.environ.get('CPU_NUM', multiprocessing.cpu_count()))
            for i in xrange(cpu_num):
108
                p = core.Place()
L
Luo Tao 已提交
109
                self._act_places.append(core.CPUPlace())
X
Xin Pan 已提交
110 111 112
                p.set_place(self._act_places[-1])
                self._places.append(p)
        assert self._places, "no place for execution"
113

Y
yuyang18 已提交
114 115
        if exec_strategy is None:
            exec_strategy = ExecutionStrategy()
116
        exec_strategy.use_cuda = use_cuda
Y
yuyang18 已提交
117 118

        if exec_strategy.num_threads == 0:
X
Xin Pan 已提交
119 120 121
            if use_cuda:
                # Experiments on se-resnext shows that too many threads hurt
                # performance. Worth tunning for other models in the future.
C
chengduoZH 已提交
122
                exec_strategy.num_threads = len(self._places) * 4
X
Xin Pan 已提交
123
            else:
C
chengduoZH 已提交
124 125 126
                cpu_num = int(
                    os.environ.get('CPU_NUM', multiprocessing.cpu_count()))
                exec_strategy.num_threads = cpu_num
127

Y
yuyang18 已提交
128 129 130
        if build_strategy is None:
            build_strategy = BuildStrategy()

131 132
        main = main_program
        main = main if main else framework.default_main_program()
133 134
        scope = executor.global_scope()

135 136 137 138 139 140
        if share_vars_from and not isinstance(share_vars_from,
                                              ParallelExecutor):
            raise TypeError("share_vars_from must be ParallelExecutor.")
        local_scopes = share_vars_from.executor.local_scopes(
        ) if share_vars_from else []

T
typhoonzero 已提交
141
        self.persistable_vars = [
142
            v.name
143 144
            for v in filter(
                lambda var: var.persistable and var.type != core.VarDesc.VarType.RAW,
T
typhoonzero 已提交
145
                main.list_vars())
146 147
        ]

148
        self.executor = core.ParallelExecutor(
X
Xin Pan 已提交
149
            self._places,
150 151 152 153
            set([
                p.name for p in main.global_block().iter_parameters()
                if not p.stop_gradient
            ]),
154 155 156
            set(self.persistable_vars), main.desc, loss_name
            if loss_name else '', scope, local_scopes, exec_strategy,
            build_strategy, num_trainers, trainer_id)
157 158
        self.scope = scope

Y
Yu Yang 已提交
159
    def run(self, fetch_list, feed=None, feed_dict=None):
X
Xin Pan 已提交
160
        """
Y
Yu Yang 已提交
161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184
        Run a parallel executor with fetch_list.

        The feed parameter can be a dict or a list. If feed is a dict, the
        feed data will be split into multiple devices. If feed is a list, we
        assume the data has been splitted into multiple devices, the each
        element in the list will be copied to each device directly.

        For example, if the feed is a dict:
        >>> exe = ParallelExecutor()
        >>> # the image will be splitted into devices. If there is two devices
        >>> # each device will process an image with shape (24, 1, 28, 28)
        >>> exe.run(feed={'image': numpy.random.random(size=(48, 1, 28, 28))})

        For example, if the feed is a list:
        >>> exe = ParallelExecutor()
        >>> # each device will process each element in the list.
        >>> # the 1st device will process an image with shape (48, 1, 28, 28)
        >>> # the 2nd device will process an image with shape (32, 1, 28, 28)
        >>> #
        >>> # you can use exe.device_count to get the device number.
        >>> exe.run(feed=[{"image": numpy.random.random(size=(48, 1, 28, 28))},
        >>>               {"image": numpy.random.random(size=(32, 1, 28, 28))},
        >>>              ])

X
Xin Pan 已提交
185

Y
Yu Yang 已提交
186 187
        Args:
            fetch_list(list): The fetched variable names
Y
Yu Yang 已提交
188 189 190 191
            feed(list|dict|None): The feed variables. If the feed is a dict,
                tensors in that dict will be splitted into each devices. If
                the feed is a list, each element of the list will be copied
                to each device.
Y
Yu Yang 已提交
192
            feed_dict: Alias for feed parameter, for backward compatibility.
Y
Yu Yang 已提交
193
                This parameter is deprecated.
Y
Yu Yang 已提交
194 195 196

        Returns: fetched result list.

X
Xin Pan 已提交
197
        """
198
        if feed is None and feed_dict is not None:
J
JiayiFeng 已提交
199
            feed = feed_dict
Y
Yu Yang 已提交
200
            print >> sys.stderr, "`feed_dict` is deprecated. Please use `feed=`"
Y
Yu Yang 已提交
201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236

        if isinstance(feed, dict):
            feed_tensor_dict = dict()
            for feed_name in feed:
                feed_tensor = feed[feed_name]
                if not isinstance(feed_tensor, core.LoDTensor):
                    feed_tensor = core.LoDTensor()
                    # always set to CPU place, since the tensor need to be splitted
                    # it is fast in CPU
                    feed_tensor.set(feed[feed_name], core.CPUPlace())
                feed_tensor_dict[feed_name] = feed_tensor

            self.executor.feed_and_split_tensor_into_local_scopes(
                feed_tensor_dict)
        elif isinstance(feed, list) or isinstance(feed, tuple):
            if len(feed) != len(self._act_places):
                raise ValueError(
                    "Feed a list of tensor, the list should be the same size as places"
                )

            res = list()

            for i, each in enumerate(feed):
                if not isinstance(each, dict):
                    raise TypeError(
                        "Each element of feed list should be a dict")
                res_dict = dict()
                for feed_name in each:
                    tensor = each[feed_name]
                    if not isinstance(tensor, core.LoDTensor):
                        tmp = core.LoDTensor()
                        tmp.set(tensor, self._act_places[i])
                        tensor = tmp
                    res_dict[feed_name] = tensor
                res.append(res_dict)
            self.executor.feed_tensors_into_local_scopes(res)
X
Xin Pan 已提交
237

238
        fetch_var_name = '@FETCHED_VAR_NAME@'
Y
Yu Yang 已提交
239
        self.executor.run(fetch_list, fetch_var_name)
240 241
        arr = self.scope.find_var(fetch_var_name).get_lod_tensor_array()
        return [arr[i] for i in range(len(arr))]
T
typhoonzero 已提交
242 243 244

    def bcast_params(self):
        self.executor.bcast_params(set(self.persistable_vars))
Y
Yu Yang 已提交
245 246 247 248

    @property
    def device_count(self):
        return len(self._act_places)