fleet_base.py 35.0 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
#   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.

from __future__ import print_function
16
import copy
17
import warnings
18
import paddle
19
from paddle.fluid.framework import dygraph_only
20
from paddle.fluid import compiler
21
from .role_maker import UserDefinedRoleMaker, PaddleCloudRoleMaker, RoleMakerBase
22
from .strategy_compiler import StrategyCompiler
23
from .distributed_strategy import DistributedStrategy
24 25
from .meta_optimizer_factory import MetaOptimizerFactory
from .runtime_factory import RuntimeFactory
26
from paddle.fluid.wrapped_decorator import wrap_decorator
27
from paddle.fluid.dygraph import parallel_helper
28

29

30 31 32 33 34 35 36 37 38 39 40 41
def _inited_runtime_handler_(func):
    def __impl__(*args, **kwargs):
        cls = args[0]

        if cls._runtime_handle is None:
            raise ValueError("Fleet can not find suitable runtime handler")

        return func(*args, **kwargs)

    return __impl__


42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57
def _is_non_distributed_check_(func):
    def __impl__(*args, **kwargs):
        cls = args[0]

        if cls._role_maker is not None and cls._role_maker._is_non_distributed(
        ) is True:
            warnings.warn(
                "%s() function doesn't work when use non_distributed fleet." %
                (func.__name__))
            return

        return func(*args, **kwargs)

    return __impl__


58
inited_runtime_handler = wrap_decorator(_inited_runtime_handler_)
59
is_non_distributed_check = wrap_decorator(_is_non_distributed_check_)
60 61


62 63 64
class Fleet(object):
    """
    Unified API for distributed training of PaddlePaddle
65
    Please reference the https://github.com/PaddlePaddle/FleetX for details
66 67 68 69 70


    Returns:
        Fleet: A Fleet instance

71
    Example for collective training:
72 73
        .. code-block:: python

74
            import paddle.distributed.fleet as fleet
75 76 77

            fleet.init(is_collective=True)

78 79 80
            strategy = fleet.DistributedStrategy()
            optimizer = paddle.optimizer.SGD(learning_rate=0.001)
            optimizer = fleet.distributed_optimizer(optimizer, strategy=strategy)
81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96

            # do distributed training


    Example for parameter server training:

        .. code-block:: python

            import paddle.distributed.fleet as fleet

            fleet.init()

            strategy = fleet.DistributedStrategy()
            optimizer = paddle.optimizer.SGD(learning_rate=0.001)
            optimizer = fleet.distributed_optimizer(optimizer, strategy=strategy)

97 98
            if fleet.is_first_worker():
                print("this is first worker")
99

100 101
            print("current node index: {}".format(fleet.worker_index()))
            print("total number of worker num: {}".format(fleet.worker_num()))
102

103 104 105
            if fleet.is_worker():
                print("this is worker")
            print("worker endpoints: {}".format(fleet.worker_endpoints(to_string=True)))
106

107 108
            print("server num: {}".format(fleet.server_num()))
            print("server endpoints: {}".format(fleet.server_endpoints(to_string=True)))
109

110 111 112
            if fleet.is_server():
                print("this is server")
            fleet.stop_worker()
113 114


115 116 117
    """

    def __init__(self):
118
        self._role_maker = None
119
        self.strategy_compiler = None
120
        self._is_collective = False
121
        self._runtime_handle = None
D
Dong Daxiang 已提交
122 123
        self._util = None
        self._context = {}
124

125 126 127 128
    def init(self, role_maker=None, is_collective=False):
        """
        Initialize role_maker in Fleet.

129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163
        This function is responsible for the distributed architecture
        what you want to run your code behind.

        Args:
            role_maker (RoleMakerBase, optional): A ``RoleMakerBase`` containing the configuration
                of environment variables related to distributed training.If you did not initialize 
                the rolemaker by yourself, it will be automatically initialized to PaddleRoleMaker.
                The default value is None.
            is_collective (Boolean, optional): A ``Boolean`` variable determines whether the program 
                runs on the CPU or GPU. False means set distributed training using CPU, and True means
                GPU.The default value is False.The default value is False.
        Returns:
            None

        Examples1:

            .. code-block:: python

                import paddle.distributed.fleet as fleet
                fleet.init()

        Examples2:

            .. code-block:: python

                import paddle.distributed.fleet as fleet
                fleet.init(is_collective=True)

        Examples3:

            .. code-block:: python

                import paddle.distributed.fleet as fleet
                role = fleet.PaddleCloudRoleMaker
                fleet.init(role)
164

165
        """
166 167

        if role_maker is None:
168 169 170 171 172 173
            if isinstance(is_collective, bool):
                self._is_collective = is_collective
                self._role_maker = PaddleCloudRoleMaker(
                    is_collective=self._is_collective)
            else:
                raise ValueError(
174 175
                    "`is_collective` should be instance of `bool`, but got {}".
                    format(type(is_collective)))
176
        else:
177 178 179 180 181 182
            if isinstance(role_maker, RoleMakerBase):
                self._role_maker = role_maker
            else:
                raise ValueError(
                    "`role_maker` should be subclass of `RoleMakerBase`, but got {}".
                    format(type(role_maker)))
183
        self._role_maker._generate_role()
184

185 186 187
        import paddle.distributed.fleet as fleet
        fleet.util._set_role_maker(self._role_maker)

188
        self.strategy_compiler = StrategyCompiler()
189 190 191 192 193 194 195 196 197

        if self._role_maker._is_non_distributed() and self._is_collective:
            if paddle.fluid.core.is_compiled_with_cuda():
                gpus_num = paddle.fluid.core.get_cuda_device_count()
                if gpus_num != 1:
                    raise ValueError(
                        "CUDA_VISIBLE_DEVICES shoule be set only 1 card if you use `python` to launch fleet program."
                    )

198
        if paddle.fluid.framework.in_dygraph_mode():
199 200
            if self.worker_num() == 1:
                return
201 202 203 204 205
            if parallel_helper._is_parallel_ctx_initialized():
                warnings.warn(
                    "The dygraph parallel environment has been initialized.")
            else:
                paddle.distributed.init_parallel_env()
206 207 208 209 210 211 212 213

    def is_first_worker(self):
        """
        Check whether the node is the first instance of worker.

        Returns:
            bool: True if this is the first node of worker,
                  False if not.
214

215 216 217 218 219 220 221 222
        Examples:

            .. code-block:: python

                import paddle.distributed.fleet as fleet
                fleet.init()
                fleet.is_first_worker()

223
        """
224
        return self._role_maker._is_first_worker()
225 226 227 228 229 230 231

    def worker_index(self):
        """
        Get current worker index.

        Returns:
            int: node id
232 233 234 235 236 237 238 239

        Examples:

            .. code-block:: python
                import paddle.distributed.fleet as fleet
                fleet.init()
                fleet.worker_index()

240
        """
241
        return self._role_maker._worker_index()
242 243 244 245 246 247 248

    def worker_num(self):
        """
        Get current total worker number.

        Returns:
            int: worker numbers
D
Dong Daxiang 已提交
249
        
250 251 252 253 254 255 256
        Examples:
            .. code-block:: python

                import paddle.distributed.fleet as fleet
                fleet.init()
                fleet.worker_num()

257
        """
258
        return self._role_maker._worker_num()
259 260 261 262 263 264 265 266

    def is_worker(self):
        """
        Check whether the node is an instance of worker.

        Returns:
            bool: True if this is a node of worker,
                  False if not.
267 268 269 270 271 272 273 274

        Examples:
            .. code-block:: python

                import paddle.distributed.fleet as fleet
                fleet.init()
                fleet.is_worker()

275
        """
276
        return self._role_maker._is_worker()
277 278 279

    def worker_endpoints(self, to_string=False):
        """
280
        Get current worker endpoints, such as ["127.0.0.1:1001", "127.0.0.1:1002"].
281 282 283

        Returns:
            list/string: server endpoints
284 285 286 287 288 289 290 291

        Examples:
            .. code-block:: python

                import paddle.distributed.fleet as fleet
                fleet.init()
                fleet.worker_endpoints()

292 293
        """
        if to_string:
294
            return ",".join(self._role_maker._get_trainer_endpoints())
295
        else:
296
            return self._role_maker._get_trainer_endpoints()
297 298 299 300 301 302 303

    def server_num(self):
        """
        Get current total worker number.

        Returns:
            int: server number
304 305 306 307 308 309

        Examples:
            .. code-block:: python
            import paddle.distributed.fleet as fleet
            fleet.init()
            fleet.server_num()
310
        """
311
        return len(self._role_maker._get_pserver_endpoints())
312 313 314 315 316 317 318

    def server_index(self):
        """
        Get current server index.

        Returns:
            int: node id
319 320 321 322 323 324 325 326

        Examples:
            .. code-block:: python

                import paddle.distributed.fleet as fleet
                fleet.init()
                fleet.server_index()

327
        """
328
        return self._role_maker._server_index()
329 330 331 332 333 334 335

    def server_endpoints(self, to_string=False):
        """
        Get current server endpoints, such as ["127.0.0.1:1001", "127.0.0.1:1002"].

        Returns:
            list/string: server endpoints
336 337 338 339 340 341 342 343

        Examples:
            .. code-block:: python

                import paddle.distributed.fleet as fleet
                fleet.init()
                fleet.server_endpoints()

344
        """
345

346
        if to_string:
347
            return ",".join(self._role_maker._get_pserver_endpoints())
348
        else:
349
            return self._role_maker._get_pserver_endpoints()
350 351 352 353 354 355 356 357

    def is_server(self):
        """
        Check whether the node is an instance of server.

        Returns:
            bool: True if this is a node of server,
                  False if not.
358 359 360 361 362 363 364 365

        Examples:

            .. code-block:: python
                import paddle.distributed.fleet as fleet
                fleet.init()
                fleet.is_server()

366
        """
367
        return self._role_maker._is_server(
368
        ) or self._role_maker._is_heter_worker()
369 370 371

    def barrier_worker(self):
        """
372 373 374 375
        barrier all workers

        Returns:
            None
376
        """
377
        self._role_maker._barrier("worker")
378

379
    @is_non_distributed_check
380
    @inited_runtime_handler
381 382
    def init_worker(self):
        """
383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400
        initialize `Communicator` for parameter server training.


        Returns:
            None

        Examples:

            .. code-block:: python

                import paddle.distributed.fleet as fleet
                fleet.init()

                # build net
                # fleet.distributed_optimizer(...)

                fleet.init_worker()

401 402 403
        """
        self._runtime_handle._init_worker()

404
    @is_non_distributed_check
405
    @inited_runtime_handler
406
    def init_server(self, *args, **kwargs):
407
        """
408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426
        init_server executor to initialize startup program,
        if the `args` is not empty, it will run load_persistables for increment training.


        Returns:
            None

        Examples:

            .. code-block:: python

                import paddle.distributed.fleet as fleet
                fleet.init()

                # build net
                # fleet.distributed_optimizer(...)

                fleet.init_server()

427
        """
428
        self._runtime_handle._init_server(*args, **kwargs)
429

430
    @is_non_distributed_check
431
    @inited_runtime_handler
432 433
    def run_server(self):
        """
434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451
        run server will run pserver main program with executor.

        Returns:
            None

        Examples:

            .. code-block:: python

                import paddle.distributed.fleet as fleet
                fleet.init()

                # build net
                # fleet.distributed_optimizer(...)

                if fleet.is_server():
                    fleet.init_server()

452 453 454
        """
        self._runtime_handle._run_server()

455
    @is_non_distributed_check
456
    @inited_runtime_handler
457 458
    def stop_worker(self):
        """
459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475
        stop `Communicator` and give training complete notice to parameter server.

        Returns:
            None

        Examples:

            .. code-block:: python

                import paddle.distributed.fleet as fleet
                fleet.init()

                # build net
                # fleet.distributed_optimizer(...)

                fleet.init_server()

476 477 478
        """
        self._runtime_handle._stop_worker()

479 480 481 482 483 484 485
    def save_inference_model(self,
                             executor,
                             dirname,
                             feeded_var_names,
                             target_vars,
                             main_program=None,
                             export_for_deployment=True):
486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505
        """
        save inference model for inference.

        Returns:
            None

        Examples:

            .. code-block:: python

                import paddle.distributed.fleet as fleet
                fleet.init()

                # build net
                # fleet.distributed_optimizer(...)

                fleet.init_server()

        """

506 507 508 509 510
        self._runtime_handle._save_inference_model(
            executor, dirname, feeded_var_names, target_vars, main_program,
            export_for_deployment)

    def save_persistables(self, executor, dirname, main_program=None):
511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550
        """

        saves all persistable variables from :code:`main_program` to
        the folder :code:`dirname`. You can refer to

        The :code:`dirname` is used to specify the folder where persistable variables
        are going to be saved. If you would like to save variables in separate
        files, set :code:`filename` None.

        Args:
            executor(Executor): The executor to run for saving persistable variables.
                                You can refer to :ref:`api_guide_executor_en` for
                                more details.

            dirname(str, optional): The saving directory path.
                                When you need to save the parameter to the memory, set it to None.
            main_program(Program, optional): The program whose persistbale variables will
                                             be saved. Default: None.


        Returns:
            None

        Examples:

            .. code-block:: text

                import paddle.distributed.fleet as fleet
                import paddle.fluid as fluid

                fleet.init()

                # build net
                # fleet.distributed_optimizer(...)

                exe = fluid.Executor(fluid.CPUPlace())
                fleet.save_persistables(exe, "dirname", fluid.default_main_program())

        """

551 552
        self._runtime_handle._save_persistables(executor, dirname, main_program)

553
    def distributed_optimizer(self, optimizer, strategy=None):
554
        """
555 556 557 558 559 560 561 562 563
        Optimizer for distributed training.

        For the distributed training, this method would rebuild a new instance of DistributedOptimizer.
        Which has basic Optimizer function and special features for distributed training.

        Args:
            optimizer(Optimizer): The executor to run for init server.
            strategy(DistributedStrategy): Extra properties for distributed optimizer.

564
        Returns:
565
            Fleet: instance of fleet.
566 567

        Examples:
568

569
            .. code-block:: python
570 571 572 573 574 575 576 577

                import paddle.distributed.fleet as fleet
                role = fleet.role_maker.PaddleCloudRoleMaker(is_collective=True)
                fleet.init(role)
                strategy = fleet.DistributedStrategy()
                optimizer = paddle.optimizer.SGD(learning_rate=0.001)
                optimizer = fleet.distributed_optimizer(optimizer, strategy=strategy)

578 579
        """
        self.user_defined_optimizer = optimizer
580

581 582
        if strategy == None:
            strategy = DistributedStrategy()
D
Dong Daxiang 已提交
583 584 585

        self._user_defined_strategy = copy.deepcopy(strategy)
        self._context = {}
586 587
        return self

588 589 590
    @dygraph_only
    def distributed_model(self, model):
        """
591 592 593 594 595 596 597
        Return distributed data parallel model (Only work in dygraph mode)

        Args:
            model (Layer): the user-defind model which inherits Layer.

        Returns:
            distributed data parallel model which inherits Layer.
598 599

        Examples:
600

601 602
            .. code-block:: python

603 604 605 606 607 608 609 610 611
                import paddle
                import paddle.nn as nn
                from paddle.distributed import fleet

                class LinearNet(nn.Layer):
                    def __init__(self):
                        super(LinearNet, self).__init__()
                        self._linear1 = nn.Linear(10, 10)
                        self._linear2 = nn.Linear(10, 1)
612

613 614
                    def forward(self, x):
                        return self._linear2(self._linear1(x))
615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644

                # 1. enable dynamic mode
                paddle.disable_static()

                # 2. initialize fleet environment
                fleet.init(is_collective=True)

                # 3. create layer & optimizer
                layer = LinearNet()
                loss_fn = nn.MSELoss()
                adam = paddle.optimizer.Adam(
                    learning_rate=0.001, parameters=layer.parameters())

                # 4. get data_parallel model using fleet
                adam = fleet.distributed_optimizer(adam)
                dp_layer = fleet.distributed_model(layer)

                # 5. run layer
                inputs = paddle.randn([10, 10], 'float32')
                outputs = dp_layer(inputs)
                labels = paddle.randn([10, 1], 'float32')
                loss = loss_fn(outputs, labels)

                print("loss:", loss.numpy())

                loss.backward()

                adam.step()
                adam.clear_grad()

645

646 647 648 649 650 651 652 653 654
        """
        assert model is not None
        self.model = paddle.DataParallel(model)
        return self.model

    @dygraph_only
    def state_dict(self):
        """
        Get state dict information from optimizer.
655
        (Only work in dygraph mode)
656 657 658 659 660 661 662

        Returns: 
            state_dict(dict) : dict contains all the Tensor used by optimizer

        Examples:
            .. code-block:: python

663 664 665 666 667 668
                import numpy as np
                import paddle
                from paddle.distributed import fleet

                paddle.disable_static()
                fleet.init(is_collective=True)
669

670 671
                value = np.arange(26).reshape(2, 13).astype("float32")
                a = paddle.fluid.dygraph.to_variable(value)
672

673 674
                layer = paddle.nn.Linear(13, 5)
                adam = paddle.optimizer.Adam(learning_rate=0.01, parameters=layer.parameters())
675

676 677 678
                adam = fleet.distributed_optimizer(adam)
                dp_layer = fleet.distributed_model(layer)
                state_dict = adam.state_dict()
679 680 681 682 683 684 685 686
        """
        # imitate target optimizer retrieval
        return self.user_defined_optimizer.state_dict()

    @dygraph_only
    def set_state_dict(self, state_dict):
        """
        Load optimizer state dict.
687
        (Only work in dygraph mode)
688 689 690 691

        Args: 
            state_dict(dict) : Dict contains all the Tensor needed by optimizer

692 693
        Returns:
            None
694 695 696 697

        Examples:
            .. code-block:: python

698 699 700
                import numpy as np
                import paddle
                from paddle.distributed import fleet
701

702 703 704 705 706
                paddle.disable_static()
                fleet.init(is_collective=True)

                value = np.arange(26).reshape(2, 13).astype("float32")
                a = paddle.fluid.dygraph.to_variable(value)
707

708 709
                layer = paddle.nn.Linear(13, 5)
                adam = paddle.optimizer.Adam(learning_rate=0.01, parameters=layer.parameters())
710

711 712 713 714 715 716
                adam = fleet.distributed_optimizer(adam)
                dp_layer = fleet.distributed_model(layer)
                state_dict = adam.state_dict()
                paddle.framework.save(state_dict, "paddle_dy")
                para_state_dict, opti_state_dict = paddle.framework.load( "paddle_dy")
                adam.set_state_dict(opti_state_dict)
717 718 719 720 721 722 723 724
        """
        # imitate target optimizer retrieval
        return self.user_defined_optimizer.set_state_dict(state_dict)

    @dygraph_only
    def set_lr(self, value):
        """
        Set the value of the learning rate manually in the optimizer. 
725
        (Only work in dygraph mode)
726

727 728 729
        Args:
            value (float|Tensor): the value of learning rate

730 731
        Returns: 
            None 
732 733 734 735

        Examples:
            .. code-block:: python

736 737 738
                import numpy as np
                import paddle
                from paddle.distributed import fleet
739

740 741
                paddle.disable_static()
                fleet.init(is_collective=True)
742

743 744
                value = np.arange(26).reshape(2, 13).astype("float32")
                a = paddle.fluid.dygraph.to_variable(value)
745

746 747
                layer = paddle.nn.Linear(13, 5)
                adam = paddle.optimizer.Adam(learning_rate=0.01, parameters=layer.parameters())
748

749 750 751 752 753 754 755 756 757 758 759 760 761 762
                adam = fleet.distributed_optimizer(adam)
                dp_layer = fleet.distributed_model(layer)

                lr_list = [0.2, 0.3, 0.4, 0.5, 0.6]
                for i in range(5):
                    adam.set_lr(lr_list[i])
                    lr = adam.get_lr()
                    print("current lr is {}".format(lr))
                # Print:
                #    current lr is 0.2
                #    current lr is 0.3
                #    current lr is 0.4
                #    current lr is 0.5
                #    current lr is 0.6
763 764 765 766 767 768 769 770
        """
        # imitate target optimizer retrieval
        return self.user_defined_optimizer.set_lr(value)

    @dygraph_only
    def get_lr(self):
        """
        Get current step learning rate.
771
        (Only work in dygraph mode)
772 773 774 775 776 777 778

        Returns:
            float: The learning rate of the current step.

        Examples:
            .. code-block:: python

779 780 781 782 783 784
                import numpy as np
                import paddle
                from paddle.distributed import fleet

                paddle.disable_static()
                fleet.init(is_collective=True)
785

786 787
                value = np.arange(26).reshape(2, 13).astype("float32")
                a = paddle.fluid.dygraph.to_variable(value)
788

789 790
                layer = paddle.nn.Linear(13, 5)
                adam = paddle.optimizer.Adam(learning_rate=0.01, parameters=layer.parameters())
791

792 793
                adam = fleet.distributed_optimizer(adam)
                dp_layer = fleet.distributed_model(layer)
794

795 796
                lr = adam.get_lr()
                print(lr) # 0.01
797 798 799 800 801 802 803 804
        """
        # imitate target optimizer retrieval
        return self.user_defined_optimizer.get_lr()

    @dygraph_only
    def step(self):
        """
        Execute the optimizer once.
805
        (Only work in dygraph mode)
806

807 808
        Returns:
            None
809 810 811 812

        Examples:
            .. code-block:: python

813 814 815
                import paddle
                import paddle.nn as nn
                from paddle.distributed import fleet
816

817 818 819 820 821
                class LinearNet(nn.Layer):
                    def __init__(self):
                        super(LinearNet, self).__init__()
                        self._linear1 = nn.Linear(10, 10)
                        self._linear2 = nn.Linear(10, 1)
822

823 824
                    def forward(self, x):
                        return self._linear2(self._linear1(x))
825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862

                # 1. enable dynamic mode
                paddle.disable_static()

                # 2. initialize fleet environment
                fleet.init(is_collective=True)

                # 3. create layer & optimizer
                layer = LinearNet()
                loss_fn = nn.MSELoss()
                adam = paddle.optimizer.Adam(
                    learning_rate=0.001, parameters=layer.parameters())

                # 4. get data_parallel model using fleet
                adam = fleet.distributed_optimizer(adam)
                dp_layer = fleet.distributed_model(layer)

                # 5. run layer
                inputs = paddle.randn([10, 10], 'float32')
                outputs = dp_layer(inputs)
                labels = paddle.randn([10, 1], 'float32')
                loss = loss_fn(outputs, labels)

                print("loss:", loss.numpy())

                loss.backward()

                adam.step()
                adam.clear_grad()


        """
        # imitate target optimizer retrieval
        return self.user_defined_optimizer.step()

    @dygraph_only
    def clear_grad(self):
        """
863 864
        Clear the gradients of all optimized parameters for model.
        (Only work in dygraph mode)
865

866 867
        Returns: 
            None
868 869 870 871

        Examples:
            .. code-block:: python

872 873 874
                import paddle
                import paddle.nn as nn
                from paddle.distributed import fleet
875

876 877 878 879 880
                class LinearNet(nn.Layer):
                    def __init__(self):
                        super(LinearNet, self).__init__()
                        self._linear1 = nn.Linear(10, 10)
                        self._linear2 = nn.Linear(10, 1)
881

882 883
                    def forward(self, x):
                        return self._linear2(self._linear1(x))
884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917

                # 1. enable dynamic mode
                paddle.disable_static()

                # 2. initialize fleet environment
                fleet.init(is_collective=True)

                # 3. create layer & optimizer
                layer = LinearNet()
                loss_fn = nn.MSELoss()
                adam = paddle.optimizer.Adam(
                    learning_rate=0.001, parameters=layer.parameters())

                # 4. get data_parallel model using fleet
                adam = fleet.distributed_optimizer(adam)
                dp_layer = fleet.distributed_model(layer)

                # 5. run layer
                inputs = paddle.randn([10, 10], 'float32')
                outputs = dp_layer(inputs)
                labels = paddle.randn([10, 1], 'float32')
                loss = loss_fn(outputs, labels)

                print("loss:", loss.numpy())

                loss.backward()

                adam.step()
                adam.clear_grad()

        """
        # imitate target optimizer retrieval
        return self.user_defined_optimizer.clear_grad()

D
Dong Daxiang 已提交
918 919 920 921 922 923 924 925 926
    def _final_strategy(self):
        if "valid_strategy" not in self._context:
            print(
                "WARNING: You may need to call minimize function before this function is called"
            )
            return {}
        else:
            return self._context["valid_strategy"]

927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949
    def minimize(self,
                 loss,
                 startup_program=None,
                 parameter_list=None,
                 no_grad_set=None):
        """
        Add distributed operations to minimize ``loss`` by updating ``parameter_list``.

        Args:
            loss (Variable): A ``Variable`` containing the value to minimize.
            startup_program (Program, optional): :ref:`api_fluid_Program` for
                initializing parameters in ``parameter_list``. The default value
                is None, at this time :ref:`api_fluid_default_startup_program` will be used.
            parameter_list (Iterable, optional): Iterable of ``Variable`` or ``Variable.name`` to update
                to minimize ``loss``. The default value is None, at this time all parameters
                will be updated.
            no_grad_set (set, optional): Set of ``Variable``  or ``Variable.name`` that don't need
                to be updated. The default value is None.

        Returns:
            tuple: tuple (optimize_ops, params_grads), A list of operators appended
            by minimize and a list of (param, grad) variable pairs, param is
            ``Parameter``, grad is the gradient value corresponding to the parameter.
950 951
            The returned tuple can be passed to ``fetch_list`` in ``Executor.run()`` to
            indicate program pruning. If so, the program will be pruned by ``feed`` and
952 953 954
            ``fetch_list`` before run, see details in ``Executor``.

        Examples:
955
            .. code-block:: python
956

957 958
                import paddle
                import paddle.distributed.fleet as fleet
959

960 961 962 963 964 965 966 967 968 969 970 971
                fc_1 = paddle.fluid.layers.fc(input=input_x, size=hid_dim, act='tanh')
                fc_2 = paddle.fluid.layers.fc(input=fc_1, size=hid_dim, act='tanh')
                prediction = paddle.fluid.layers.fc(input=[fc_2], size=label_dim, act='softmax')
                cost = paddle.fluid.layers.cross_entropy(input=prediction, label=input_y)
                avg_cost = paddle.fluid.layers.mean(x=cost)

                role = fleet.role_maker.PaddleCloudRoleMaker(is_collective=True)
                fleet.init(role)
                strategy = fleet.DistributedStrategy()
                optimizer = paddle.optimizer.SGD(learning_rate=0.001)
                optimizer = fleet.distributed_optimizer(optimizer, strategy=strategy)
                optimizer.minimize(avg_cost)
972

973
                # for more examples, please reference https://github.com/PaddlePaddle/FleetX
974 975

        """
D
Dong Daxiang 已提交
976 977 978
        context = {}
        context["user_defined_strategy"] = copy.deepcopy(
            self._user_defined_strategy)
979 980 981
        if paddle.fluid.framework.in_dygraph_mode():
            # imitate target optimizer retrieval
            target_opt = self.user_defined_optimizer
D
Dong Daxiang 已提交
982
            self._context = context
983 984
            return target_opt.minimize(loss)

985 986
        # cache original feed forward program
        self.origin_main_program = loss.block.program
987 988
        context["origin_main_program"] = self.origin_main_program
        context["loss"] = loss
989 990
        if startup_program == None:
            self.origin_startup_program = \
991 992
                paddle.static.default_startup_program().clone(for_test=False)
            startup_program = paddle.static.default_startup_program()
993 994 995
        else:
            self.origin_startup_program = \
                startup_program.clone(for_test=False)
996

997 998
        context["origin_startup_program"] = startup_program
        context["role_maker"] = self._role_maker
999 1000 1001 1002 1003

        # compile time
        distributed_optimizer_list = \
            MetaOptimizerFactory()._get_valid_meta_optimizers(
                self.user_defined_optimizer)
D
Dong Daxiang 已提交
1004

D
Dong Daxiang 已提交
1005 1006 1007
        context["user_defined_strategy"] = copy.deepcopy(
            self._user_defined_strategy)
        copy_user_defined_strategy = copy.deepcopy(self._user_defined_strategy)
1008 1009 1010 1011 1012 1013

        # trigger the auto-parallel in very strict condition
        # strategy = DistributedStrategy()
        # strategy.auto = True
        # optimizer = paddle.optimizer.SGD(learning_rate=0.1)
        # optimizer = fleet.distributed_optimizer(optimizer, strategy)
D
Dong Daxiang 已提交
1014
        if copy_user_defined_strategy._is_strict_auto():
1015 1016
            # turn on all the strategy for each optimizer
            for opt in distributed_optimizer_list:
D
Dong Daxiang 已提交
1017
                opt._enable_strategy(copy_user_defined_strategy, context)
1018

1019 1020
        valid_optimizer_list = []
        valid_graph_optimizer_list = []
D
Dong Daxiang 已提交
1021
        can_not_apply_optimizer_list = []
1022 1023 1024 1025
        # recall meta optimizers for ranking
        for opt in distributed_optimizer_list:
            opt._set_basic_info(loss, self._role_maker,
                                self.user_defined_optimizer,
D
Dong Daxiang 已提交
1026
                                copy_user_defined_strategy)
1027 1028
            if opt._can_apply() and not opt._is_graph_out():
                valid_optimizer_list.append(opt)
D
Dong Daxiang 已提交
1029
            elif opt._can_apply() and opt._is_graph_out():
1030
                valid_graph_optimizer_list.append(opt)
D
Dong Daxiang 已提交
1031 1032
            else:
                can_not_apply_optimizer_list.append(opt)
1033
        # combine recalled meta optimizers to be a valid meta optimizer
D
Dong Daxiang 已提交
1034
        meta_optimizer, graph_optimizer = \
1035 1036
            self.strategy_compiler.generate_optimizer(
                loss, self._role_maker, self.user_defined_optimizer,
D
Dong Daxiang 已提交
1037
                copy_user_defined_strategy, valid_optimizer_list,
1038
                valid_graph_optimizer_list)
D
Dong Daxiang 已提交
1039

D
Dong Daxiang 已提交
1040
        valid_strategy = self.strategy_compiler._get_valid_strategy(
D
Dong Daxiang 已提交
1041 1042 1043
            copy_user_defined_strategy, can_not_apply_optimizer_list)

        context["valid_strategy"] = copy.deepcopy(valid_strategy)
1044

D
Dong Daxiang 已提交
1045
        self._context = context
1046

D
Dong Daxiang 已提交
1047
        self.valid_strategy = valid_strategy
1048
        self.valid_strategy._enable_env()
D
Dong Daxiang 已提交
1049

1050 1051
        optimize_ops = []
        params_grads = []
1052

1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
        if self._role_maker._is_non_distributed() and not self._is_collective:
            if self._runtime_handle is None:
                self._runtime_handle = RuntimeFactory()._create_runtime(context)

            compiled_program = compiler.CompiledProgram(
                self.origin_main_program).with_data_parallel(
                    loss_name=loss.name, share_vars_from=None)
            loss.block.program._graph = compiled_program
            return self.user_defined_optimizer.minimize(
                loss,
                startup_program=startup_program,
                parameter_list=parameter_list,
                no_grad_set=no_grad_set)

1067 1068 1069 1070 1071 1072
        if meta_optimizer:
            optimize_ops, params_grads = meta_optimizer.minimize(
                loss,
                startup_program=startup_program,
                parameter_list=parameter_list,
                no_grad_set=no_grad_set)
1073

1074
            default_program = paddle.static.default_main_program()
1075 1076 1077 1078

            if id(default_program) != id(loss.block.program):
                paddle.fluid.framework.switch_main_program(loss.block.program)

1079 1080 1081 1082 1083 1084
        else:
            optimize_ops, params_grads = self.user_defined_optimizer.minimize(
                loss,
                startup_program=startup_program,
                parameter_list=parameter_list,
                no_grad_set=no_grad_set)
1085

1086 1087
        context["program_optimize_ops"] = optimize_ops
        context["program_params_grads"] = params_grads
1088

1089
        if graph_optimizer:
D
Dong Daxiang 已提交
1090
            optimize_ops, params_grads = graph_optimizer.minimize(
1091 1092 1093 1094 1095 1096 1097 1098
                loss,
                startup_program=startup_program,
                parameter_list=parameter_list,
                no_grad_set=no_grad_set)
            # since we do not encourage users to use graph operations
            # if a graph optimizer takes effect, mostly
            # optimizers_ops and params_grads are None
            # i.e. users can not modify current computation graph anymore
1099 1100 1101
            context["graph_optimize_ops"] = optimize_ops
            context["graph_optimize_grads"] = params_grads

1102
        if self._runtime_handle is None:
1103
            self._runtime_handle = RuntimeFactory()._create_runtime(context)
1104

1105 1106
        import paddle.distributed.fleet as fleet
        fleet.util._set_strategy(context["valid_strategy"])
1107 1108

        return optimize_ops, params_grads