reader.py 68.9 KB
Newer Older
S
sneaxiy 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14
# Copyright (c) 2019 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.

15
from . import core
16
import sys
S
sneaxiy 已提交
17
import six
18
import numpy as np
S
sneaxiy 已提交
19
import threading
20
import paddle
21
from .framework import Program, Variable, program_guard, default_main_program, default_startup_program, in_dygraph_mode, cpu_places, _current_expected_place
S
sneaxiy 已提交
22
from .executor import global_scope
23
from .data_feeder import DataFeeder, BatchedTensorProvider
24 25 26
from .multiprocess_utils import multiprocess_queue_set, CleanupFuncRegistrar, _cleanup_mmap, _cleanup, _set_SIGCHLD_handler
from .dataloader import BatchSampler, Dataset
from .dataloader.dataloader_iter import _DataLoaderIterSingleProcess, _DataLoaderIterMultiProcess
S
sneaxiy 已提交
27
from .layers.io import monkey_patch_reader_methods, _copy_reader_var_, double_buffer
S
sneaxiy 已提交
28
from .unique_name import UniqueNameGenerator
29
import logging
30
import warnings
Z
Zeng Jinle 已提交
31
from .dataset import DatasetBase, InMemoryDataset
S
sneaxiy 已提交
32

33
### Dygraph DataLoader configs ###
34
import os
35 36
import multiprocessing
import signal
37

38
# NOTE: queue has a different name in python2 and python3
39
if six.PY2:
40 41 42
    import Queue as queue
else:
    import queue
43

44 45 46
# NOTE: [ avoid hanging & failed quickly ] These value is used in getting data from another process
QUEUE_GET_TIMEOUT = 60

Z
Zeng Jinle 已提交
47 48 49
__all__ = ['PyReader', 'DataLoader']

data_loader_unique_name_generator = UniqueNameGenerator()
S
sneaxiy 已提交
50

51 52 53 54 55 56 57 58 59 60 61
KEEP_DATA_LOADER_ORDER = True


def keep_data_loader_order(*args):
    global KEEP_DATA_LOADER_ORDER
    if len(args) == 0:
        return KEEP_DATA_LOADER_ORDER
    else:
        assert len(args) == 1 and isinstance(args[0], bool)
        KEEP_DATA_LOADER_ORDER = args[0]

S
sneaxiy 已提交
62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77

def _convert_places(places):
    if not isinstance(places, (list, tuple)):
        places = [places]

    ret = []
    for p in places:
        if not isinstance(p, core.Place):
            tmp = core.Place()
            tmp.set_place(p)
            p = tmp

        ret.append(p)
    return ret


Z
Zeng Jinle 已提交
78 79 80
class DataLoaderBase(object):
    def __init__(self):
        self._places = None
S
sneaxiy 已提交
81

Z
Zeng Jinle 已提交
82 83
    def __call__(self):
        return self
S
sneaxiy 已提交
84

Z
Zeng Jinle 已提交
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
    def next(self):
        '''
        Get the next item in the DataLoader object. This method    
        should not be called by users directly. It is used for
        implementing iterator protocol of Python 2.x inside
        PaddlePaddle framework.
        '''
        return self.__next__()

    def __iter__(self):
        raise NotImplementedError()

    def __next__(self):
        raise NotImplementedError()

100 101 102 103 104 105 106 107 108 109 110 111
    @classmethod
    def _check_input_array(cls, item):
        arr = np.asarray(item)
        if arr.dtype == np.object:
            raise TypeError(
                "\n\tFaild to convert input data to a regular ndarray :\n\t* Usually "
                "this means the input data contains nested lists with different lengths. "
                "\n\t* Check the reader function passed to 'decorate_batch_generator'"
                " to locate the data causes this issue.\n\t* Please consider using "
                "'fluid.create_lod_tensor' to convert it to a LoD-Tensor.")
        return arr

Z
Zeng Jinle 已提交
112 113

class DataLoader(object):
114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328
    """
    DataLoader prodives an iterator which iterates given dataset
    once by the batch_sampler.

    DataLoader supports single-process and multi-prcess data loading,
    multi-process workers will be used to load data asynchronously if
    :attr:`num_workers` is set as a positive number.

    DataLoader only supports map-style dataset(can get a sample from
    dataset with a given index) currently, for a map-style dataset,
    please see :code:`paddle.io.Dataset`.

    batch_sampler please see :code:`paddle.io.BatchSampler`

    Args:  
        dataset(Dataset): the dataset to load data from, should be an
            instance of subclass of :code:`paddle.io.Dataset`.
        feed_list (list(Variable)|tuple(Variable)): feed variable list.
            The variables should be created by :code:`fluid.data()`.
            :attr:`feed_list` must be set if :attr:`return_list` is
            False. Default None.
        places(list(Place)|tuple(Place)): a list of Place, to put data
            onto, :attr:`places` must be set in both static graph and 
            dynamic graph mode, in dynamic graph mode, place number must
            be 1. Default None.
        return_list (bool): whether the return value on each device is 
            presented as a list. If :attr:`return_list=False`, the return
            value on each device would be a dict of str -> LoDTensor, where
            the key of the dict is the name of each fed variables. If 
            :attr:`return_list=True`, the return value on each device would
            be a list(LoDTensor). :attr:`return_list` can only be True
            in dynamic graph mode. Default False.
        batch_sampler(BatchSampler): an instance of `paddle.io.BatchSampler`
            to generate batch indices to draw samples from :attr:`dataset`
            and combine a batch. Default None.
        batch_size(int): sample number in a mini-batch, a substitution
            parameter for :attr:`batch_sampler`, if :attr:`batch_sampler`
            is not set, a default `paddle.io.BatchSampler` will be used
            and initialize by :attr:`batch_size`, :attr:`shuffle` and
            :attr:`drop_last`. Default 1.
        shuffle(bool): whther to shuffle indices order before genrate
            batch indices, a substitution parameter for :attr:`batch_sampler`
            see :attr:`batch_size`. Default False.
        drop_last(bool): whether drop the last incomplete batch dataset size
            is not divisible by the batch size, a substitution parameter
            for :attr:`batch_sampler`, see :attr:`batch_size`. Default False
        collate_fn(callable): function to generate mini-batch data by merging
            the sample list, None for only stack each fields of sample in axis
            0(same as :attr::`np.stack(..., axis=0)`). Default None
        num_workers(int): the number of subprocess to load data, 0 for no
            subprocess used and loading data in main process. Default 0
        use_buffer_reader (bool): whether to use bufferred reader. 
            If use_buffer_reader=True, the DataLoader would prefetch next 
            batch data asynchronously, so it would speed up data feeding 
            and occupies a little more CPU or GPU memory, i.e., the memory
            of one batch input data. Default True.
        use_shared_memory (bool): whether to use shared memory to speed up
            putting data into inter-process queue, set :attr:`use_shared_memory`
            as True only when the shared memory space on your machine(e.g.
            space of '/dev/shm' on Linux operating sysytem) is large enough.
            Shared memory will only be enabled in multi-process mode(num_workers
            > 0). Default True.
        timeout(int): the timeout value for getting data form output queue
            of subprocesses. Default 0.
        worker_init_fn(callable): init function which will be called with
            worker id on each subproces starting if not set as None. Default
            None.

    Returns:
        DataLoader: an iterable object for data iterating

    Examples:
        
        .. code-block:: python

            import numpy as np
            import paddle.fluid as fluid
            from paddle.io import Dataset, BatchSampler, DataLoader

            BATCH_NUM = 20
            BATCH_SIZE = 16
            EPOCH_NUM = 4

            IMAGE_SIZE = 784
            CLASS_NUM = 10

            USE_GPU = False # whether use GPU to run model

            # define a random dataset
            class RandomDataset(Dataset):
                def __init__(self, num_samples):
                    self.num_samples = num_samples

                def __getitem__(self, idx):
                    image = np.random.random([IMAGE_SIZE]).astype('float32')
                    label = np.random.randint(0, CLASS_NUM - 1, (1, )).astype('int64')
                    return image, label

                def __len__(self):
                    return self.num_samples

            # get places
            places = fluid.cuda_places() if USE_GPU else fluid.cpu_places()

            # -------------------- static graph ---------------------

            def simple_net(image, label):
                fc_tmp = fluid.layers.fc(image, size=CLASS_NUM, act='softmax')
                cross_entropy = fluid.layers.softmax_with_cross_entropy(image, label)
                loss = fluid.layers.reduce_mean(cross_entropy)
                sgd = fluid.optimizer.SGD(learning_rate=1e-3)
                sgd.minimize(loss)
                return loss

            image = fluid.data(name='image', shape=[None, IMAGE_SIZE], dtype='float32')
            label = fluid.data(name='label', shape=[None, 1], dtype='int64')

            loss = simple_net(image, label)

            exe = fluid.Executor(places[0])
            exe.run(fluid.default_startup_program())

            prog = fluid.CompiledProgram(fluid.default_main_program()).with_data_parallel(loss_name=loss.name)

            dataset = RandomDataset(BATCH_NUM * BATCH_SIZE)

            loader = DataLoader(dataset,
                                feed_list=[image, label],
                                places=places,
                                batch_size=BATCH_SIZE, 
                                shuffle=True,
                                drop_last=True,
                                num_workers=2)

            for e in range(EPOCH_NUM):
                for i, data in enumerate(loader()):
                    l = exe.run(prog, feed=data, fetch_list=[loss], return_numpy=True)
                    print("Epoch {} batch {}: loss = {}".format(e, i, l[0][0]))

            # -------------------------------------------------------
                
            # --------------------- dygraph mode --------------------

            class SimpleNet(fluid.dygraph.Layer):
                def __init__(self):
                    super(SimpleNet, self).__init__()
                    self.fc = fluid.dygraph.nn.Linear(IMAGE_SIZE, CLASS_NUM, act='softmax')

                def forward(self, image, label=None):
                    return self.fc(image)

            with fluid.dygraph.guard(places[0]):
                simple_net = SimpleNet()
                opt = fluid.optimizer.SGD(learning_rate=1e-3,
                                          parameter_list=simple_net.parameters())

                loader = DataLoader(dataset,
                                    places=places[0],
                                    batch_size=BATCH_SIZE,
                                    shuffle=True,
                                    drop_last=True,
                                    num_workers=2)

                for e in range(EPOCH_NUM):
                    for i, (image, label) in enumerate(loader()):
                        out = simple_net(image)
                        loss = fluid.layers.cross_entropy(out, label)
                        avg_loss = fluid.layers.reduce_mean(loss)
                        avg_loss.backward()
                        opt.minimize(avg_loss)
                        simple_net.clear_gradients()
                        print("Epoch {} batch {}: loss = {}".format(e, i, np.mean(loss.numpy())))

            # -------------------------------------------------------

    """

    def __init__(self,
                 dataset,
                 feed_list=None,
                 places=None,
                 return_list=False,
                 batch_sampler=None,
                 batch_size=1,
                 shuffle=False,
                 drop_last=False,
                 collate_fn=None,
                 num_workers=0,
                 use_buffer_reader=True,
                 use_shared_memory=True,
                 timeout=0,
                 worker_init_fn=None):
        self.return_list = return_list
        self.collate_fn = collate_fn
        self.use_buffer_reader = use_buffer_reader
        self.worker_init_fn = worker_init_fn

        assert isinstance(dataset, Dataset), \
            "dataset should be subclass instance of paddle.io.Dataset"
        self.dataset = dataset

        if not return_list and not in_dygraph_mode():
            assert feed_list is not None, \
                    "feed_list should be set when return_list=False"
        self.feed_list = feed_list

        assert places is not None, "places cannot be None"
        self.places = _convert_places(places)
        if in_dygraph_mode():
            assert len(self.places) == 1, \
                    "Number of places must be 1 in dygraph mode"

        assert num_workers >= 0, "num_workers should be a non-negative value"
        if num_workers > 0 and (sys.platform == 'darwin' or
                                sys.platform == 'win32'):
329 330 331
            warnings.warn(
                "DataLoader with multi-process mode is not supported on MacOs and Windows currently." \
                " Please use signle-process mode with num_workers = 0 instead")
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371
            num_workers = 0
        self.num_workers = num_workers

        self.use_shared_memory = use_shared_memory
        if use_shared_memory and num_workers == 0:
            self.use_shared_memory = False

        assert timeout >= 0, "timeout should be a non-negative value"
        self.timeout = timeout

        if batch_sampler is not None:
            assert isinstance(batch_sampler, BatchSampler), \
                "batch_sampler should be None or subclass instance " \
                "of paddle.io.BatchSampler"
            assert batch_size == 1 and not shuffle and not drop_last, \
                "batch_size/shuffle/drop_last should not be set when " \
                "batch_sampler is given"
            self.batch_sampler = batch_sampler
        else:
            assert batch_size is not None and batch_size > 0, \
                "batch_size should be a positive value when " \
                "batch_sampler is not given"
            self.batch_sampler = BatchSampler(
                dataset=dataset,
                batch_size=batch_size,
                shuffle=shuffle,
                drop_last=drop_last)

    def __len__(self):
        return len(self.batch_sampler)

    def __iter__(self):
        if self.num_workers == 0:
            return _DataLoaderIterSingleProcess(self)
        else:
            return _DataLoaderIterMultiProcess(self)

    def __call__(self):
        return self.__iter__()

Z
Zeng Jinle 已提交
372 373 374 375 376
    @staticmethod
    def from_generator(feed_list=None,
                       capacity=None,
                       use_double_buffer=True,
                       iterable=True,
377
                       return_list=False,
378 379
                       use_multiprocess=False,
                       drop_last=True):
Z
Zeng Jinle 已提交
380
        """
381 382 383
        .. note::
          **The framework ensures that the data loading order of DataLoader is exactly the same as the user-defined data source.**

Z
Zeng Jinle 已提交
384 385 386 387 388 389 390 391
        Create a DataLoader object for loading data from Python generator. 
        Data would be prefetched using Python thread and be pushed
        into a queue asynchronously.

        The created DataLoader object provides 3 methods to set the data source
        :code:`set_sample_generator` , :code:`set_sample_list_generator` and 
        :code:`set_batch_generator` . Please see the following example codes
        to know their usages.
392
        
Z
Zeng Jinle 已提交
393 394 395 396 397 398 399 400 401 402 403 404
        If iterable = True, the created DataLoader object is a Python generator
        object, which is iterable using for-range loop.

        If iterable = False, the created DataLoader object provides 
        :code:`start()` and :code:`reset()` method to control the data reading
        process. This mode is designed to be compatible with the 
        :code:`fluid.layers.py_reader` interface. Users can migrate the codes   
        from :code:`fluid.layers.py_reader` to :code:`fluid.io.DataLoader` 
        easily when using iterable=False. 

        Args:  
            feed_list (list(Variable)|tuple(Variable)): feed variable list.
405
                The variables should be created by :code:`fluid.data()`.
Z
Zeng Jinle 已提交
406 407 408 409 410 411 412 413 414 415 416 417 418
            capacity (int): capacity of the queue maintained in DataLoader.
                The unit is batch number. Set larger capacity if your reader 
                is fast. 
            use_double_buffer (bool): whether to use double_buffer_reader. 
                If use_double_buffer=True, the DataLoader would prefetch next 
                batch data asynchronously, so it would speed up data feeding 
                and occupies a little more CPU or GPU memory, i.e., the memory
                of one batch input data. 
            iterable (bool): whether the created DataLoader is iterable. 
            return_list (bool): whether the return value on each device is 
                presented as a list. It is only valid when iterable=True. 
                If return_list=False, the return value on each device would 
                be a dict of str -> LoDTensor, where the key of the dict is 
T
tianshuo78520a 已提交
419
                the name of each fed variables. If return_list=True, the 
Z
Zeng Jinle 已提交
420 421
                return value on each device would be a list(LoDTensor). It is
                recommended to use return_list=False in static graph mode and
422 423 424 425 426 427
                use return_list=True in dygraph mode.  
            use_multiprocess (bool): whether to use multi-process to speed up
                the data loading process in dygraph. Note: this parameter only
                can be used in the dygraph mode. In the static graph mode,
                whether this parameter is set or not has no effect.
                The Default value is False.
428 429 430 431 432 433 434
            drop_last (bool): whether to drop the last batches whose number is
                less than the CPU core/GPU card number. The default value is 
                True. In training phase, users should not set drop_last=False,
                because all CPU cores/GPU cards must read data from DataLoader. 
                In inference phase, users can set drop_last=False, so that the
                last batches whose number is less than the CPU core/GPU card
                number can be tested. 
Z
Zeng Jinle 已提交
435 436 437 438

        Returns:
            loader (DataLoader): the created DataLoader object.

439
        Examples 1:
Z
Zeng Jinle 已提交
440 441
            
            .. code-block:: python
S
sneaxiy 已提交
442

Z
Zeng Jinle 已提交
443 444
                import paddle.fluid as fluid
                import numpy as np
445

Z
Zeng Jinle 已提交
446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
                BATCH_NUM = 10 
                BATCH_SIZE = 16
                EPOCH_NUM = 4

                CLASS_NUM = 10

                ITERABLE = True # whether the created DataLoader object is iterable
                USE_GPU = False # whether to use GPU

                DATA_FORMAT = 'batch_generator' # data format of data source user provides 

                def simple_net(image, label):
                    fc_tmp = fluid.layers.fc(image, size=CLASS_NUM)
                    cross_entropy = fluid.layers.softmax_with_cross_entropy(image, label)
                    loss = fluid.layers.reduce_mean(cross_entropy)
                    sgd = fluid.optimizer.SGD(learning_rate=1e-3)
                    sgd.minimize(loss)
                    return loss

                def get_random_images_and_labels(image_shape, label_shape):
                    image = np.random.random(size=image_shape).astype('float32')
                    label = np.random.random(size=label_shape).astype('int64')
                    return image, label

                # If the data generator yields one sample each time,
                # use DataLoader.set_sample_generator to set the data source.
                def sample_generator_creator(): 
                    def __reader__():
                        for _ in range(BATCH_NUM * BATCH_SIZE):
                            image, label = get_random_images_and_labels([784], [1])
                            yield image, label

                    return __reader__

                # If the data generator yield list of samples each time,
                # use DataLoader.set_sample_list_generator to set the data source.
                def sample_list_generator_creator():
                    def __reader__():
                        for _ in range(BATCH_NUM): 
                            sample_list = []
                            for _ in range(BATCH_SIZE):
                                image, label = get_random_images_and_labels([784], [1])
                                sample_list.append([image, label])

                            yield sample_list

                    return __reader__ 

                # If the data generator yields a batch each time, 
                # use DataLoader.set_batch_generator to set the data source.
                def batch_generator_creator():
                    def __reader__():
                        for _ in range(BATCH_NUM):
                            batch_image, batch_label = get_random_images_and_labels([BATCH_SIZE, 784], [BATCH_SIZE, 1]) 
                            yield batch_image, batch_label
H
Huihuang Zheng 已提交
501

Z
Zeng Jinle 已提交
502
                    return __reader__
503

Z
Zeng Jinle 已提交
504 505 506 507 508
                # If DataLoader is iterable, use for loop to train the network 
                def train_iterable(exe, prog, loss, loader):
                    for _ in range(EPOCH_NUM):
                        for data in loader():
                            exe.run(prog, feed=data, fetch_list=[loss])
509

Z
Zeng Jinle 已提交
510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528
                # If DataLoader is not iterable, use start() and reset() method to control the process 
                def train_non_iterable(exe, prog, loss, loader):
                    for _ in range(EPOCH_NUM):
                        loader.start() # call DataLoader.start() before each epoch starts
                        try:
                            while True:
                                exe.run(prog, fetch_list=[loss])
                        except fluid.core.EOFException:
                            loader.reset() # call DataLoader.reset() after catching EOFException 

                def set_data_source(loader, places):
                    if DATA_FORMAT == 'sample_generator':
                        loader.set_sample_generator(sample_generator_creator(), batch_size=BATCH_SIZE, drop_last=True, places=places)
                    elif DATA_FORMAT == 'sample_list_generator':
                        loader.set_sample_list_generator(sample_list_generator_creator(), places=places)
                    elif DATA_FORMAT == 'batch_generator':
                        loader.set_batch_generator(batch_generator_creator(), places=places)
                    else:
                        raise ValueError('Unsupported data format')
529

530 531
                image = fluid.data(name='image', shape=[None, 784], dtype='float32')
                label = fluid.data(name='label', shape=[None, 1], dtype='int64')
532

Z
Zeng Jinle 已提交
533 534
                # Define DataLoader 
                loader = fluid.io.DataLoader.from_generator(feed_list=[image, label], capacity=16, iterable=ITERABLE)
535

Z
Zeng Jinle 已提交
536 537
                # Define network
                loss = simple_net(image, label)
S
sneaxiy 已提交
538

Z
Zeng Jinle 已提交
539 540 541 542 543 544 545 546 547
                # Set data source of DataLoader
                #
                # If DataLoader is iterable, places must be given and the number of places must be the same with device number.  
                #  - If you are using GPU, call `fluid.cuda_places()` to get all GPU places. 
                #  - If you are using CPU, call `fluid.cpu_places()` to get all CPU places. 
                # 
                # If DataLoader is not iterable, places can be None.
                places = fluid.cuda_places() if USE_GPU else fluid.cpu_places()
                set_data_source(loader, places)
S
sneaxiy 已提交
548

Z
Zeng Jinle 已提交
549 550
                exe = fluid.Executor(places[0])
                exe.run(fluid.default_startup_program())
H
Huihuang Zheng 已提交
551

Z
Zeng Jinle 已提交
552
                prog = fluid.CompiledProgram(fluid.default_main_program()).with_data_parallel(loss_name=loss.name)
553

Z
Zeng Jinle 已提交
554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570
                if loader.iterable:
                    train_iterable(exe, prog, loss, loader)
                else:
                    train_non_iterable(exe, prog, loss, loader)


                '''
                Users can use return_list = True in dygraph mode. 
                '''
                with fluid.dygraph.guard(places[0]):
                    loader = fluid.io.DataLoader.from_generator(capacity=2, return_list=True)
                    set_data_source(loader, places[0]) 
                    for image, label in loader():
                        relu = fluid.layers.relu(image)
                        assert image.shape == [BATCH_SIZE, 784] 
                        assert label.shape == [BATCH_SIZE, 1]
                        assert relu.shape == [BATCH_SIZE, 784]
571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613

        Examples 2:

            .. code-block:: python

                import paddle.fluid as fluid
                import numpy as np
                import os

                # We use 2 CPU cores to run inference network 
                os.environ['CPU_NUM'] = '2'

                # The data source has only 3 batches, which can not be
                # divided evenly to each CPU core
                def batch_generator():  
                    for i in range(3):
                        yield np.array([i+1]).astype('float32'), 

                x = fluid.data(name='x', shape=[None], dtype='float32')  
                y = x * x

                def run_inference(drop_last): 
                    loader = fluid.io.DataLoader.from_generator(feed_list=[x],
                            capacity=8, drop_last=drop_last)
                    loader.set_batch_generator(batch_generator, fluid.cpu_places())

                    exe = fluid.Executor(fluid.CPUPlace())
                    prog = fluid.CompiledProgram(fluid.default_main_program())
                    prog = prog.with_data_parallel()

                    result = []
                    for data in loader():
                        each_ret, = exe.run(prog, feed=data, fetch_list=[y])
                        result.extend(each_ret)
                    return result

                # Set drop_last to True, so that the last batch whose
                # number is less than CPU core number would be discarded.
                print(run_inference(drop_last=True)) # [1.0, 4.0]

                # Set drop_last to False, so that the last batch whose
                # number is less than CPU core number can be tested.
                print(run_inference(drop_last=False)) # [1.0, 4.0, 9.0]
Z
Zeng Jinle 已提交
614
        """
615 616 617 618 619 620
        if in_dygraph_mode():
            return DygraphGeneratorLoader(feed_list, capacity,
                                          use_double_buffer, iterable,
                                          return_list, use_multiprocess)
        else:
            return GeneratorLoader(feed_list, capacity, use_double_buffer,
621
                                   iterable, return_list, drop_last)
Z
Zeng Jinle 已提交
622 623 624 625 626 627

    @staticmethod
    def from_dataset(dataset, places, drop_last=True):
        """
        Create an iterable DataLoader object for loading data from Dataset.    
        Dataset is only supported in Linux system currently.
628

Z
Zeng Jinle 已提交
629 630 631 632 633 634 635
        Args:
            dataset (InMemoryDataset|QueueDataset): the dataset object.
            places (list(CUDAPlace)|list(CPUPlace)): places where the result 
                data should be converted.   
            drop_last (bool): whether to drop the last batch whose sample 
                number is less than batch size. If drop_last = True, they
                would be dropped. If drop_last = False, they would be kept. 
636

Z
Zeng Jinle 已提交
637 638 639
        Returns:
            loader (DataLoader): the created DataLoader object, which can be 
                treated as a Python generator.   
640

Z
Zeng Jinle 已提交
641 642 643
        Examples:

            .. code-block:: python
644

Z
Zeng Jinle 已提交
645
                import paddle.fluid as fluid
646

647 648
                image = fluid.data(name='image', shape=[None, 784], dtype='float32')
                label = fluid.data(name='label', shape=[None, 1], dtype='int64')
649

Z
Zeng Jinle 已提交
650 651 652 653 654
                dataset = fluid.DatasetFactory().create_dataset("QueueDataset")
                dataset.set_batch_size(32)
                dataset.set_filelist(['a.txt', 'b.txt', 'c.txt'])
                dataset.set_use_var([image, label])
                dataset.set_pipe_command('cat') 
655

Z
Zeng Jinle 已提交
656 657 658
                loader = fluid.io.DataLoader.from_dataset(dataset, fluid.cpu_places())
        """
        return DatasetLoader(dataset, places, drop_last)
S
sneaxiy 已提交
659

S
sneaxiy 已提交
660

661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685
class DygraphGeneratorLoader(DataLoaderBase):
    """
    The GeneratorLoader of dygraph

    The multiprocess dygraph GeneratorLoader's most functions are different from 
    static graph GeneratorLoader, Separate implementation to keep code readable.
    """

    def __init__(self,
                 feed_list=None,
                 capacity=None,
                 use_double_buffer=True,
                 iterable=True,
                 return_list=True,
                 use_multiprocess=False):
        self._batch_reader = None
        self._places = None
        self._feed_list = feed_list

        if not capacity:
            raise ValueError("Please give value to capacity.")
        self._capacity = capacity
        self._use_double_buffer = use_double_buffer

        if not iterable:
686 687
            warnings.warn(
                "Please NOTE: DygraphGeneratorLoader supports iterable mode only. Change to iterable mode."
688 689 690
            )
        self._iterable = True
        if not return_list:
691 692
            warnings.warn(
                "Please NOTE: DygraphGeneratorLoader supports returning as list only. Change to return as list."
693 694 695 696 697 698 699
            )
        self._return_list = True

        # NOTE: the multiprocessing in different platform is incompatible, we will solve it later
        self._use_multiprocess = use_multiprocess
        if self._use_multiprocess and (sys.platform == 'darwin' or
                                       sys.platform == 'win32'):
700 701
            warnings.warn(
                "NOTE: DygraphGeneratorLoader with multiprocess mode is not currently supported on MacOs and Windows."
702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726
            )
            self._use_multiprocess = False

        if self._use_multiprocess:
            # NOTE: the multiprocessing.Queue used to save loading data in self._process
            self._data_queue = None
            # NOTE: this process is used to load data asynchronously from self._batch_reader
            self._process = None

        # NOTE: the C++ LoDTensorBlockingQueue instance
        self._blocking_queue = None
        # NOTE: 1. In multiprocess mode, this thread is used to get next batch data from
        # self._data_queue, then push it into self._blocking_queue; 2. In singleprocess
        # mode, this thread is used to get next batch data from self._batch_reader, then 
        # push it into self._blocking_queue
        self._thread = None

    @property
    def queue(self):
        return self._blocking_queue

    @property
    def iterable(self):
        return self._iterable

727 728 729 730 731 732 733 734 735 736
    def _clear_and_remove_data_queue(self):
        if self._data_queue is not None:
            while True:
                try:
                    self._data_queue.get_nowait()
                except queue.Empty:
                    break
            global multiprocess_queue_set
            multiprocess_queue_set.remove(self._data_queue)

737 738 739 740 741 742 743 744 745 746 747
    def _wait_thread_ends(self):
        thread = self._thread
        if thread is not None:
            self._blocking_queue.close()
            thread.join()

    def _wait_process_ends(self):
        process = self._process
        if process is not None:
            process.join()
            # erase process id
748
            core._erase_process_pids(id(self))
749

750 751 752 753 754 755 756 757 758
    def _init_iterable(self):
        self._wait_thread_ends()
        if self._use_multiprocess:
            self._wait_process_ends()
        self._var_names = []
        self._shapes = []
        self._dtypes = []
        self._need_check_feed = []
        self._blocking_queue = core.init_lod_tensor_blocking_queue(
759
            core.Variable(), self._capacity, False)
760
        self._reader = None
761 762
        self._reader = core.create_py_reader(
            self.queue, self._var_names, self._shapes, self._dtypes,
763
            self._need_check_feed, self._places, self._use_double_buffer, True)
764 765 766

    def _start(self):
        if self._use_multiprocess:
767 768 769
            # clear old _data_queue and remove it from multiprocess_queue_set
            self._clear_and_remove_data_queue()
            # set data_queue and process
770
            self._data_queue = multiprocessing.Queue(self._capacity)
771 772 773
            # add _data_queue into global queue set
            global multiprocess_queue_set
            multiprocess_queue_set.add(self._data_queue)
774 775 776 777 778 779 780 781 782 783 784
            self._process = multiprocessing.Process(
                target=self._reader_process_loop)
            self._process.daemon = True
            self._process.start()

            # Set child process signal handler
            # NOTE: [ avoiding hang ] 1. if the child process dies due to bus error/segfault
            # or just hang, the main process will hang waiting for data, so here need to deal 
            # with SIGSEGV and SIGBUS of child process; 2. if the main process end before child
            # process, it shuts the all its daemonic children down with a SIGTERM (instead of 
            # joining them without a timeout), so here nedd to deal with SIGTERM.
785 786
            core._set_process_pids(id(self), [self._process.pid])
            _set_SIGCHLD_handler()
787 788 789 790

            # Set reader_thread
            self._thread_done_event = threading.Event()
            self._thread = threading.Thread(
791
                target=self._reader_thread_loop_for_multiprocess)
792 793 794
            self._thread.daemon = True
            self._thread.start()
        else:
795 796
            self._thread = threading.Thread(
                target=self._reader_thread_loop_for_singleprocess)
797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821
            self._thread.daemon = True
            self._thread.start()

    def _reset(self):
        self._reader.reset()
        self._wait_thread_ends()
        if self._use_multiprocess:
            self._wait_process_ends()

    def __iter__(self):
        assert self.iterable, "DataLoader is not iterable"
        assert self._batch_reader is not None, \
            "Data source of DataLoader has not set yet"

        self._init_iterable()
        self._start()
        return self

    def __next__(self):
        try:
            return self._reader.read_next_var_list()
        except StopIteration:
            self._reset()
            six.reraise(*sys.exc_info())

822 823 824 825 826 827 828 829 830
    def _exit_thread_expectedly(self):
        self._thread_done_event.set()
        self._blocking_queue.close()

    def _exit_thread_unexpectedly(self):
        self._thread_done_event.set()
        self._blocking_queue.kill()
        logging.error("DataLoader reader thread raised an exception!")

831 832 833 834 835
    def _reader_process_loop(self):
        try:
            # set signal handler
            core._set_process_signal_handler()

836 837 838 839
            # NOTE: [ mmap files clear ] When the child process exits unexpectedly,
            # some shared memory objects may have been applied for but have not yet
            # been put into the inter-process Queue. This part of the object needs
            # to be cleaned up when the process ends.
840
            CleanupFuncRegistrar.register(_cleanup_mmap)
841 842 843 844 845

            for batch in self._batch_reader():
                tensor_list = core._convert_to_tensor_list(batch)
                self._data_queue.put(tensor_list)
                core._remove_tensor_list_mmap_fds(tensor_list)
846 847 848 849 850 851 852
            self._data_queue.put(None)
        except KeyboardInterrupt:
            # NOTE: Main process will raise KeyboardInterrupt anyways, ignore it in child process
            pass
        except:
            six.reraise(*sys.exc_info())

853
    def _reader_thread_loop_for_multiprocess(self):
854 855 856 857 858 859 860
        while not self._thread_done_event.is_set():
            try:
                # NOTE: [ avoid hanging ] Even with carefully designed data dependencies 
                # (i.e., a put() always corresponding to a get()), hanging on get() can 
                # still happen when data in queue is corrupted (e.g., due to 
                # Queue.cancel_join_thread or unexpected exit). So we set a timeout whenever 
                # we try to get data from `data_queue`
861 862 863 864 865 866 867
                # NOTE: [ avoid failed quickly ] Here, the time setting of QUEUE_GET_TIMEOUT
                # is relatively long, currently it is 60 seconds, because in some models,
                # if the reader child process starts with a heavy burden, the child process
                # has no enough time to put the data in the queue when the main process
                # start trying to get data from queue. At this time, the child thread needs
                # to wait slightly longer
                tensor_list = self._data_queue.get(timeout=QUEUE_GET_TIMEOUT)
868 869 870 871
            except:
                # NOTE [ avoid handing ] After adding the shared memory mechanism, not only
                # the queue.Empty exception will occur here, but other exceptions will also
                # occur, such as mmap failure. If it is not handled here, it will hang.
872
                self._exit_thread_unexpectedly()
873 874
                logging.error(
                    "DataLoader reader thread failed to read data from the multiprocessing.Queue."
875
                )
876
                six.reraise(*sys.exc_info())
877 878

            if not self._thread_done_event.is_set():
879
                if tensor_list is not None:
880 881
                    try:
                        array = core.LoDTensorArray()
882 883
                        for tensor in tensor_list:
                            array.append(tensor)
884 885 886
                        if not self._blocking_queue.push(array):
                            self._blocking_queue.close()
                    except:
887
                        self._exit_thread_unexpectedly()
888 889
                        six.reraise(*sys.exc_info())
                else:
890
                    self._exit_thread_expectedly()
891

892
    def _reader_thread_loop_for_singleprocess(self):
893 894 895 896 897
        try:
            for sample in self._batch_reader():
                array = core.LoDTensorArray()
                for item in sample:
                    if not isinstance(item, core.LoDTensor):
898
                        item = self._check_input_array(item)
899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945
                        tmp = core.LoDTensor()
                        tmp.set(item, core.CPUPlace())
                        item = tmp

                    array.append(item)

                if not self._blocking_queue.push(array):
                    break

            self._blocking_queue.close()
            self._thread = None
        except Exception:
            self._blocking_queue.kill()
            self._thread = None
            logging.warning(
                "DygraphDataLoader reader thread raised an exception.")
            six.reraise(*sys.exc_info())

    def set_sample_generator(self,
                             reader,
                             batch_size,
                             drop_last=True,
                             places=None):
        assert batch_size > 0, "batch_size must be larger than 0"
        self.set_sample_list_generator(
            paddle.batch(
                reader, batch_size=batch_size, drop_last=drop_last),
            places=places)
        return self

    def set_sample_list_generator(self, reader, places=None):
        def __batch_reader_impl__():
            for batch in reader():
                slots = []
                for items in batch:
                    for i, item in enumerate(items):
                        if len(slots) < len(items):
                            slots.append([item])
                        else:
                            slots[i].append(item)
                yield slots

        self.set_batch_generator(__batch_reader_impl__, places)
        return self

    def set_batch_generator(self, reader, places=None):
        self._batch_reader = reader
946 947
        if places is None:
            places = _current_expected_place()
948 949
        self._places = _convert_places(places)
        assert len(self._places) == 1, \
950
            "Number of places must be 1 in imperative mode"
951 952 953
        return self


Z
Zeng Jinle 已提交
954
class GeneratorLoader(DataLoaderBase):
S
sneaxiy 已提交
955
    def __init__(self,
956 957
                 feed_list=None,
                 capacity=None,
S
sneaxiy 已提交
958
                 use_double_buffer=True,
959
                 iterable=True,
960 961
                 return_list=False,
                 drop_last=True):
S
sneaxiy 已提交
962
        self._tensor_reader = None
Z
Zeng Jinle 已提交
963
        self._places = None
S
sneaxiy 已提交
964
        self._thread = None
965
        self._queue = None
966
        self._feed_list = feed_list
967 968 969
        self._exited = False
        self._drop_last = drop_last
        self._keep_order = keep_data_loader_order()
970 971
        if not capacity:
            raise ValueError("Please give value to capacity.")
972 973 974 975
        self._iterable = iterable
        self._return_list = return_list
        if not self._feed_list:
            raise Exception("Feed list must be given under static mode.")
S
sneaxiy 已提交
976 977 978 979
        self._use_double_buffer = use_double_buffer
        self._capacity = capacity
        if not self._iterable:
            self._init_non_iterable()
S
sneaxiy 已提交
980

Z
Zeng Jinle 已提交
981
    def _wait_thread_ends(self):
982
        # Get self._thread first to prevent data race, because __thread_main__
Z
Zeng Jinle 已提交
983 984 985 986 987 988 989 990
        # would set self._thread be None at the end
        thread = self._thread
        if thread is not None and self._iterable:
            self._queue.close()
            thread.join()

    def _init_iterable(self):
        self._wait_thread_ends()
991 992 993 994 995 996
        self._var_names = [v.name for v in self._feed_list]
        self._shapes = [v.shape for v in self._feed_list]
        self._dtypes = [v.dtype for v in self._feed_list]
        self._need_check_feed = [
            v.desc.need_check_feed() for v in self._feed_list
        ]
997 998
        self._queue = core.init_lod_tensor_blocking_queue(
            core.Variable(), self._capacity, self._keep_order)
999
        self._reader = None
S
sneaxiy 已提交
1000
        self._reader = core.create_py_reader(
1001
            self.queue, self._var_names, self._shapes, self._dtypes,
1002 1003
            self._need_check_feed, self._places, self._use_double_buffer,
            self._drop_last)
S
sneaxiy 已提交
1004 1005 1006 1007 1008 1009 1010

    def _init_non_iterable(self):
        lod_levels = []
        dtypes = []
        shape_concat = []
        ranks = []
        shapes = []
1011
        need_check_feed = []
S
sneaxiy 已提交
1012 1013 1014 1015 1016 1017 1018

        for feed_data in self._feed_list:
            dtypes.append(feed_data.dtype)
            shape_concat.extend(feed_data.shape)
            ranks.append(len(feed_data.shape))
            shapes.append(feed_data.shape)
            lod_levels.append(feed_data.lod_level)
1019
            need_check_feed.append(int(feed_data.desc.need_check_feed()))
S
sneaxiy 已提交
1020

Z
Zeng Jinle 已提交
1021 1022 1023 1024
        queue_name = data_loader_unique_name_generator(
            'lod_tensor_blocking_queue')
        reader_name = data_loader_unique_name_generator('create_py_reader')
        double_buffer_name = data_loader_unique_name_generator('double_buffer')
S
sneaxiy 已提交
1025

S
sneaxiy 已提交
1026
        var = global_scope().var(queue_name)
1027 1028 1029 1030 1031 1032 1033
        self._queue = core.init_lod_tensor_blocking_queue(var, self._capacity,
                                                          self._keep_order)

        if self._keep_order:
            block = default_main_program().current_block()
        else:
            block = default_startup_program().current_block()
S
sneaxiy 已提交
1034

1035
        reader_var = block.create_var(name=reader_name)
S
sneaxiy 已提交
1036

1037
        dtype_int = [int(t) for t in dtypes]
1038
        block.append_op(
S
sneaxiy 已提交
1039 1040
            type='create_py_reader',
            inputs={'blocking_queue': [queue_name]},
1041
            outputs={'Out': [reader_var]},
S
sneaxiy 已提交
1042 1043 1044
            attrs={
                'shape_concat': shape_concat,
                'lod_levels': lod_levels,
1045 1046
                'dtypes': dtype_int,
                'need_check_feed': need_check_feed,
S
sneaxiy 已提交
1047 1048 1049
                'ranks': ranks
            })

1050 1051 1052
        reader_var.desc.set_dtypes(dtypes)
        reader_var.persistable = True
        reader_var.stop_gradient = True
S
sneaxiy 已提交
1053

1054 1055 1056 1057 1058 1059 1060 1061 1062 1063
        if self._keep_order:
            main_prog_var = reader_var
            reader = main_prog_var
            reader.reset = self._queue.reset
        else:
            main_prog_var = _copy_reader_var_(
                default_main_program().current_block(), reader_var)

            main_prog_var.stop_gradient = True
            main_prog_var.persistable = True
S
sneaxiy 已提交
1064

1065
            reader = monkey_patch_reader_methods(main_prog_var)
S
sneaxiy 已提交
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079

        if self._use_double_buffer:
            double_buffer_reader = double_buffer(
                reader, name=double_buffer_name)
            # we return a double buffer reader. However, the reset method comes from
            # py_reader.
            double_buffer_reader.reset = reader.reset
            reader = double_buffer_reader

        self._reader = reader

        default_main_program().current_block().append_op(
            type='read',
            inputs={'Reader': [self._reader]},
1080 1081
            outputs={'Out': self._feed_list},
            attrs={'drop_last': self._drop_last})
S
sneaxiy 已提交
1082 1083 1084 1085 1086 1087 1088 1089

    @property
    def queue(self):
        return self._queue

    @property
    def iterable(self):
        return self._iterable
S
sneaxiy 已提交
1090

Z
Zeng Jinle 已提交
1091 1092
    def __iter__(self):
        assert self.iterable, "DataLoader is not iterable"
S
sneaxiy 已提交
1093
        assert self._tensor_reader is not None, \
Z
Zeng Jinle 已提交
1094
            "Data source of DataLoader has not set yet"
S
sneaxiy 已提交
1095

Z
Zeng Jinle 已提交
1096
        self._init_iterable()
S
sneaxiy 已提交
1097
        self._start()
Z
Zeng Jinle 已提交
1098 1099 1100 1101
        return self

    def __next__(self):
        try:
1102 1103
            if self._return_list:
                return self._reader.read_next_list()
1104
            else:
1105
                return self._reader.read_next()
Z
Zeng Jinle 已提交
1106 1107 1108 1109 1110 1111
        except StopIteration:
            self._queue.close()
            self._reset()
            six.reraise(*sys.exc_info())

    def start(self):
1112 1113
        assert not self._iterable, "start() cannot be called when DataLoader is iterable"
        self._start()
Z
Zeng Jinle 已提交
1114 1115

    def reset(self):
1116 1117
        assert not self._iterable, "reset() cannot be called when DataLoader is iterable"
        self._reset()
Z
Zeng Jinle 已提交
1118 1119 1120 1121

    def _start(self):
        def __thread_main__():
            try:
1122 1123 1124 1125
                while not self._queue.wait_for_inited(1):
                    if self._exited:
                        return

Z
Zeng Jinle 已提交
1126 1127 1128 1129
                for tensors in self._tensor_reader():
                    array = core.LoDTensorArray()
                    for item in tensors:
                        if not isinstance(item, core.LoDTensor):
1130
                            item = self._check_input_array(item)
Z
Zeng Jinle 已提交
1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
                            tmp = core.LoDTensor()
                            tmp.set(item, core.CPUPlace())
                            item = tmp

                        array.append(item)

                    if not self._queue.push(array):
                        break

                self._queue.close()
                self._thread = None
            except Exception as ex:
Z
Zeng Jinle 已提交
1143
                self._queue.kill()
Z
Zeng Jinle 已提交
1144 1145 1146 1147 1148 1149 1150
                self._thread = None
                logging.warn('Your reader has raised an exception!')
                six.reraise(*sys.exc_info())

        self._thread = threading.Thread(target=__thread_main__)
        self._thread.daemon = True
        self._thread.start()
S
sneaxiy 已提交
1151

S
sneaxiy 已提交
1152
    def _reset(self):
1153
        self._queue.close()
1154
        self._exited = True
Z
Zeng Jinle 已提交
1155 1156 1157 1158
        thread = self._thread
        if thread is not None:
            thread.join()

1159
        self._exited = False
1160 1161
        self._reader.reset()

Z
Zeng Jinle 已提交
1162 1163 1164 1165 1166 1167
    def set_sample_generator(self,
                             reader,
                             batch_size,
                             drop_last=True,
                             places=None):
        assert batch_size > 0, "batch_size must be larger than 0"
1168 1169 1170 1171 1172 1173 1174
        has_lod = False
        for f in self._feed_list:
            if f.lod_level != 0:
                has_lod = True
                break

        if has_lod:
1175 1176 1177 1178 1179
            self.set_sample_list_generator(
                paddle.batch(
                    reader, batch_size=batch_size, drop_last=drop_last),
                places=places)
        else:
1180 1181 1182 1183 1184 1185 1186
            reader = BatchedTensorProvider(
                feed_list=self._feed_list,
                place=core.CPUPlace(),
                batch_size=batch_size,
                generator=reader,
                drop_last=drop_last)
            self.set_batch_generator(reader, places=places)
Z
Zeng Jinle 已提交
1187 1188 1189
        return self

    def set_sample_list_generator(self, reader, places=None):
1190 1191 1192 1193
        with program_guard(Program(), Program()):
            feeder = DataFeeder(
                feed_list=self._feed_list, place=core.CPUPlace())
            paddle_reader = feeder.decorate_reader(reader, multi_devices=False)
Z
Zeng Jinle 已提交
1194

1195 1196 1197
        def __tensor_reader_impl__():
            for slots in paddle_reader():
                yield [slots[var.name] for var in self._feed_list]
Z
Zeng Jinle 已提交
1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236

        self.set_batch_generator(__tensor_reader_impl__, places)
        return self

    def set_batch_generator(self, reader, places=None):
        self._tensor_reader = reader
        if self._iterable:
            assert places is not None, "Places cannot be None when DataLoader is iterable"
            self._places = _convert_places(places)
        else:
            if places is not None:
                logging.info(
                    'places would be ommited when DataLoader is not iterable')
        return self


class PyReader(DataLoaderBase):
    """
    Create a reader object for data feeding in Python. 
    Data would be prefetched using Python thread and be pushed
    into a queue asynchronously. Data in the queue would be extracted 
    automatically when `Executor.run(...)` is called.

    Args:  
        feed_list (list(Variable)|tuple(Variable)): feed variable list.
            The variables should be created by :code:`fluid.layers.data()`.
        capacity (int): capacity of the queue maintained in PyReader.
            The unit is batch number. Set larger capacity if your reader 
            is fast. 
        use_double_buffer (bool): whether to use double_buffer_reader. 
            If use_double_buffer=True, PyReader would prefetch next 
            batch data asynchronously, so it would speed up data feeding 
            and occupies a little more CPU or GPU memory, i.e., the memory
            of one batch input data. 
        iterable (bool): whether the created PyReader is iterable. 
        return_list (bool): whether the return value on each device is 
            presented as a list. It is only valid when iterable=True. 
            If return_list=False, the return value on each device would 
            be a dict of str -> LoDTensor, where the key of the dict is 
T
tianshuo78520a 已提交
1237
            the name of each fed variables. If return_list=True, the 
Z
Zeng Jinle 已提交
1238 1239 1240 1241 1242
            return value on each device would be a list(LoDTensor). It is
            recommended to use return_list=False in static graph mode and
            use return_list=True in dygraph mode. 

    Returns:
G
guofei 已提交
1243 1244 1245 1246
        the created reader object.

    Return type:
        reader(Reader)
Z
Zeng Jinle 已提交
1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265

    Examples:
        1. If iterable = False, the created PyReader object is almost the
           same as :code:`fluid.layers.py_reader()`. Operators would be 
           inserted into the program. User should call :code:`start()` 
           before each epoch and catch :code:`fluid.core.EOFException`
           thrown by :code:`Executor.run()` when epoch ends. Once the 
           exception is caught, user should call :code:`reset()` to reset 
           the reader manually.

        .. code-block:: python

           import paddle
           import paddle.fluid as fluid
           import numpy as np

           EPOCH_NUM = 3
           ITER_NUM = 5
           BATCH_SIZE = 3
G
guofei 已提交
1266 1267 1268 1269 1270
           
           def network(image, label):
               # User-defined network, here is an example of softmax regression.
               predict = fluid.layers.fc(input=image, size=10, act='softmax')           
               return fluid.layers.cross_entropy(input=predict, label=label)
Z
Zeng Jinle 已提交
1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281

           def reader_creator_random_image_and_label(height, width):
               def reader():
                   for i in range(ITER_NUM):
                       fake_image = np.random.uniform(low=0,
                                                      high=255,
                                                      size=[height, width])
                       fake_label = np.ones([1])
                       yield fake_image, fake_label
               return reader

G
guofei 已提交
1282 1283
           image = fluid.data(name='image', shape=[None, 784, 784], dtype='float32')
           label = fluid.data(name='label', shape=[None, 1], dtype='int64')
Z
Zeng Jinle 已提交
1284 1285 1286 1287 1288 1289 1290 1291

           reader = fluid.io.PyReader(feed_list=[image, label],
                                      capacity=4,
                                      iterable=False)

           user_defined_reader = reader_creator_random_image_and_label(784, 784)
           reader.decorate_sample_list_generator(
               paddle.batch(user_defined_reader, batch_size=BATCH_SIZE))
G
guofei 已提交
1292 1293
           loss = network(image, label)
           executor = fluid.Executor(fluid.CPUPlace())
Z
Zeng Jinle 已提交
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320
           executor.run(fluid.default_startup_program())
           for i in range(EPOCH_NUM):
               reader.start()
               while True:
                   try:
                       executor.run(feed=None)
                   except fluid.core.EOFException:
                       reader.reset()
                       break

 
        2. If iterable=True, the created PyReader object is decoupled with
           the program. No operator would be inserted into the program. 
           In this case, the created reader is a Python generator, which 
           is iterable. User should feed the data yielded from PyReader 
           object into :code:`Executor.run(feed=...)`.  

        .. code-block:: python

           import paddle
           import paddle.fluid as fluid
           import numpy as np

           EPOCH_NUM = 3
           ITER_NUM = 5
           BATCH_SIZE = 10

G
guofei 已提交
1321 1322 1323 1324 1325
           def network(image, label):
               # User-defined network, here is an example of softmax regression.
               predict = fluid.layers.fc(input=image, size=10, act='softmax')           
               return fluid.layers.cross_entropy(input=predict, label=label)

Z
Zeng Jinle 已提交
1326 1327 1328
           def reader_creator_random_image(height, width):
               def reader():
                   for i in range(ITER_NUM):
G
guofei 已提交
1329 1330 1331
                       fake_image = np.random.uniform(low=0, high=255, size=[height, width])
                       fake_label = np.ones([1])
                       yield fake_image, fake_label 
Z
Zeng Jinle 已提交
1332 1333
               return reader

G
guofei 已提交
1334 1335 1336
           image = fluid.data(name='image', shape=[None, 784, 784], dtype='float32')
           label = fluid.data(name='label', shape=[None, 1], dtype='int64')
           reader = fluid.io.PyReader(feed_list=[image, label], capacity=4, iterable=True, return_list=False)
Z
Zeng Jinle 已提交
1337 1338 1339 1340

           user_defined_reader = reader_creator_random_image(784, 784)
           reader.decorate_sample_list_generator(
               paddle.batch(user_defined_reader, batch_size=BATCH_SIZE),
G
guofei 已提交
1341 1342 1343 1344 1345 1346
                   fluid.core.CPUPlace())
           
           loss = network(image, label)
           executor = fluid.Executor(fluid.CPUPlace())
           executor.run(fluid.default_startup_program())
           
Z
Zeng Jinle 已提交
1347 1348
           for _ in range(EPOCH_NUM):
               for data in reader():
G
guofei 已提交
1349
                   executor.run(feed=data, fetch_list=[loss])
Z
Zeng Jinle 已提交
1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403


        3. If return_list=True, the return values would be presented as list instead of dict. 
           This is usually used in dygraph mode.

        .. code-block:: python

           import paddle
           import paddle.fluid as fluid
           import numpy as np

           ITER_NUM = 5
           BATCH_SIZE = 10

           def reader_creator_random_image(height, width):
               def reader():
                   for i in range(ITER_NUM):
                       yield np.random.uniform(low=0, high=255, size=[height, width]), \
                           np.random.random_integers(low=0, high=9, size=[1])
               return reader

           place = fluid.CPUPlace()
           with fluid.dygraph.guard(place):
               py_reader = fluid.io.PyReader(capacity=2, return_list=True)
               user_defined_reader = reader_creator_random_image(784, 784)
               py_reader.decorate_sample_list_generator(
                   paddle.batch(user_defined_reader, batch_size=BATCH_SIZE),
                   place)
               for image, label in py_reader():
                   relu = fluid.layers.relu(image)
    """

    def __init__(self,
                 feed_list=None,
                 capacity=None,
                 use_double_buffer=True,
                 iterable=True,
                 return_list=False):
        self._loader = DataLoader.from_generator(
            feed_list, capacity, use_double_buffer, iterable, return_list)

    @property
    def queue(self):
        return self._loader.queue

    @property
    def iterable(self):
        return self._loader.iterable

    def __iter__(self):
        return self._loader.__iter__()

    def __next__(self):
        return self._loader.__next__()
S
sneaxiy 已提交
1404 1405

    def start(self):
S
add doc  
sneaxiy 已提交
1406 1407 1408
        '''
        Start the data feeding thread. 
        Can only call when the reader object is not iterable.  
1409
        
G
guofei 已提交
1410 1411
	Example:
	    .. code-block:: python
Z
Zeng Jinle 已提交
1412
    
H
Huihuang Zheng 已提交
1413 1414 1415 1416
                import paddle
                import paddle.fluid as fluid
                import numpy as np

1417 1418 1419 1420 1421 1422
                BATCH_SIZE = 10

                def generator():
                    for i in range(5):
                        yield np.random.uniform(low=0, high=255, size=[784, 784]),

G
guofei 已提交
1423
                image = fluid.data(name='image', shape=[None, 784, 784], dtype='float32')
1424 1425 1426 1427
                reader = fluid.io.PyReader(feed_list=[image], capacity=4, iterable=False)
                reader.decorate_sample_list_generator(
                    paddle.batch(generator, batch_size=BATCH_SIZE))

G
guofei 已提交
1428
                executor = fluid.Executor(fluid.CPUPlace())
1429 1430 1431 1432 1433 1434 1435 1436 1437 1438
                executor.run(fluid.default_startup_program())
                for i in range(3):
                    reader.start()
                    while True:
                        try:
                            executor.run(feed=None)
                        except fluid.core.EOFException:
                            reader.reset()
                            break

Z
Zeng Jinle 已提交
1439 1440
	    '''
        self._loader.start()
S
sneaxiy 已提交
1441

S
sneaxiy 已提交
1442
    def reset(self):
S
add doc  
sneaxiy 已提交
1443 1444 1445
        '''
        Reset the reader object when :code:`fluid.core.EOFException` raises. 
        Can only call when the reader object is not iterable.
1446 1447 1448 1449
        
        Example:
            .. code-block:: python

H
Huihuang Zheng 已提交
1450 1451 1452 1453
                import paddle
                import paddle.fluid as fluid
                import numpy as np

1454 1455 1456 1457 1458 1459
                BATCH_SIZE = 10

                def generator():
                    for i in range(5):
                        yield np.random.uniform(low=0, high=255, size=[784, 784]),

G
guofei 已提交
1460
                image = fluid.data(name='image', shape=[None, 784, 784], dtype='float32')
1461 1462 1463 1464
                reader = fluid.io.PyReader(feed_list=[image], capacity=4, iterable=False)
                reader.decorate_sample_list_generator(
                    paddle.batch(generator, batch_size=BATCH_SIZE))

G
guofei 已提交
1465
                executor = fluid.Executor(fluid.CPUPlace())
1466 1467 1468 1469 1470 1471 1472 1473 1474 1475
                executor.run(fluid.default_startup_program())
                for i in range(3):
                    reader.start()
                    while True:
                        try:
                            executor.run(feed=None)
                        except fluid.core.EOFException:
                            reader.reset()
                            break        

S
add doc  
sneaxiy 已提交
1476
        '''
Z
Zeng Jinle 已提交
1477
        self._loader.reset()
S
sneaxiy 已提交
1478

S
sneaxiy 已提交
1479 1480 1481 1482 1483 1484 1485 1486 1487
    def decorate_sample_generator(self,
                                  sample_generator,
                                  batch_size,
                                  drop_last=True,
                                  places=None):
        '''
        Set the data source of the PyReader object.
        
        The provided :code:`sample_generator` should be a Python generator,
1488
        which yields list(numpy.ndarray)-typed data of each sample.
S
sneaxiy 已提交
1489 1490 1491 1492

        :code:`places` must be set when the PyReader object is iterable.

        If all inputs have no lods, this method is faster than 
S
sneaxiy 已提交
1493
        :code:`decorate_sample_list_generator(paddle.batch(sample_generator, ...))` .
S
sneaxiy 已提交
1494 1495 1496

        Args:
            sample_generator (generator): Python generator that yields
1497
                list(numpy.ndarray)-typed sample data.
S
sneaxiy 已提交
1498 1499 1500 1501 1502
            batch_size (int): batch size. Must be larger than 0.
            drop_last (bool): Whether to drop the last batch when sample number
                is less than batch_size. 
            places (None|list(CUDAPlace)|list(CPUPlace)): place list. Must
                be provided when PyReader is iterable.
1503 1504 1505 1506

        Example:
            .. code-block:: python

H
Huihuang Zheng 已提交
1507 1508 1509
                import paddle.fluid as fluid
                import numpy as np

1510 1511 1512
                EPOCH_NUM = 3
                ITER_NUM = 15
                BATCH_SIZE = 3
G
guofei 已提交
1513 1514 1515 1516 1517
        
                def network(image, label):
                    # User-defined network, here is an example of softmax regression.
                    predict = fluid.layers.fc(input=image, size=10, act='softmax')           
                    return fluid.layers.cross_entropy(input=predict, label=label)
1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528

                def random_image_and_label_generator(height, width):
                    def generator():
                        for i in range(ITER_NUM):
                            fake_image = np.random.uniform(low=0,
                                                           high=255,
                                                           size=[height, width])
                            fake_label = np.array([1])
                            yield fake_image, fake_label
                    return generator

G
guofei 已提交
1529 1530
                image = fluid.data(name='image', shape=[None, 784, 784], dtype='float32')
                label = fluid.data(name='label', shape=[None, 1], dtype='int64')
1531 1532 1533 1534 1535
                reader = fluid.io.PyReader(feed_list=[image, label], capacity=4, iterable=True)

                user_defined_generator = random_image_and_label_generator(784, 784)
                reader.decorate_sample_generator(user_defined_generator,
                                                 batch_size=BATCH_SIZE,
G
guofei 已提交
1536 1537 1538 1539
                                                 places=[fluid.CPUPlace()])
                loss = network(image, label)
                executor = fluid.Executor(fluid.CPUPlace())
                executor.run(fluid.default_startup_program())
1540 1541 1542

                for _ in range(EPOCH_NUM):
                    for data in reader():
G
guofei 已提交
1543
                        executor.run(feed=data, fetch_list=[loss])
1544
    
S
sneaxiy 已提交
1545
        '''
Z
Zeng Jinle 已提交
1546 1547
        self._loader.set_sample_generator(sample_generator, batch_size,
                                          drop_last, places)
S
sneaxiy 已提交
1548

S
sneaxiy 已提交
1549
    def decorate_sample_list_generator(self, reader, places=None):
S
add doc  
sneaxiy 已提交
1550 1551 1552 1553
        '''
        Set the data source of the PyReader object. 

        The provided :code:`reader` should be a Python generator,
S
sneaxiy 已提交
1554
        which yields list(numpy.ndarray) typed batched data. 
S
add doc  
sneaxiy 已提交
1555 1556 1557 1558
        
        :code:`places` must be set when the PyReader object is iterable.

        Args:
S
sneaxiy 已提交
1559 1560 1561 1562
            reader (generator): Python generator that yields 
                list(numpy.ndarray)-typed batched data. 
            places (None|list(CUDAPlace)|list(CPUPlace)): place list. Must
                be provided when PyReader is iterable.
1563 1564 1565 1566
        
        Example:
            .. code-block:: python

H
Huihuang Zheng 已提交
1567 1568 1569 1570
                import paddle
                import paddle.fluid as fluid
                import numpy as np

1571 1572 1573 1574
                EPOCH_NUM = 3
                ITER_NUM = 15
                BATCH_SIZE = 3

G
guofei 已提交
1575 1576 1577 1578 1579
                def network(image, label):
                    # User-defined network, here is an example of softmax regression.
                    predict = fluid.layers.fc(input=image, size=10, act='softmax')           
                    return fluid.layers.cross_entropy(input=predict, label=label)

1580 1581 1582 1583 1584 1585 1586 1587 1588 1589
                def random_image_and_label_generator(height, width):
                    def generator():
                        for i in range(ITER_NUM):
                            fake_image = np.random.uniform(low=0,
                                                           high=255,
                                                           size=[height, width])
                            fake_label = np.ones([1])
                            yield fake_image, fake_label
                    return generator

G
guofei 已提交
1590 1591
                image = fluid.data(name='image', shape=[None, 784, 784], dtype='float32')
                label = fluid.data(name='label', shape=[None, 1], dtype='int64')
1592 1593 1594 1595 1596
                reader = fluid.io.PyReader(feed_list=[image, label], capacity=4, iterable=True)

                user_defined_generator = random_image_and_label_generator(784, 784)
                reader.decorate_sample_list_generator(
                    paddle.batch(user_defined_generator, batch_size=BATCH_SIZE),
G
guofei 已提交
1597 1598 1599 1600 1601
                    fluid.core.CPUPlace())
                
                loss = network(image, label)
                executor = fluid.Executor(fluid.core.CPUPlace())
                executor.run(fluid.default_startup_program())
1602 1603 1604

                for _ in range(EPOCH_NUM):
                    for data in reader():
G
guofei 已提交
1605
                        executor.run(feed=data, fetch_list=[loss])
1606
                 
S
add doc  
sneaxiy 已提交
1607
        '''
Z
Zeng Jinle 已提交
1608
        self._loader.set_sample_list_generator(reader, places)
S
sneaxiy 已提交
1609

S
sneaxiy 已提交
1610
    def decorate_batch_generator(self, reader, places=None):
S
add doc  
sneaxiy 已提交
1611 1612 1613 1614
        '''
        Set the data source of the PyReader object.

        The provided :code:`reader` should be a Python generator,
S
sneaxiy 已提交
1615
        which yields numpy.ndarray-typed or LoDTensor-typed batched data.
S
add doc  
sneaxiy 已提交
1616 1617 1618 1619 1620 1621

        :code:`places` must be set when the PyReader object is iterable.

        Args:
            reader (generator): Python generator that yields LoDTensor-typed
                batched data.
S
sneaxiy 已提交
1622
            places (None|list(CUDAPlace)|list(CPUPlace)): place list. Must
S
sneaxiy 已提交
1623
                be provided when PyReader is iterable.
1624 1625 1626 1627

        Example:
            .. code-block:: python

H
Huihuang Zheng 已提交
1628 1629 1630
                import paddle.fluid as fluid
                import numpy as np

1631 1632 1633
                EPOCH_NUM = 3
                ITER_NUM = 15
                BATCH_SIZE = 3
G
guofei 已提交
1634 1635 1636 1637 1638
               
                def network(image, label):
                    # User-defined network, here is an example of softmax regression.
                    predict = fluid.layers.fc(input=image, size=10, act='softmax')           
                    return fluid.layers.cross_entropy(input=predict, label=label)
1639 1640 1641 1642 1643 1644 1645 1646

                def random_image_and_label_generator(height, width):
                    def generator():
                        for i in range(ITER_NUM):
                            batch_image = np.random.uniform(low=0,
                                                            high=255,
                                                            size=[BATCH_SIZE, height, width])
                            batch_label = np.ones([BATCH_SIZE, 1])
G
guofei 已提交
1647 1648
                            batch_image = batch_image.astype('float32')
                            batch_label = batch_label.astype('int64')
1649 1650 1651
                            yield batch_image, batch_label
                    return generator

G
guofei 已提交
1652 1653
                image = fluid.data(name='image', shape=[None, 784, 784], dtype='float32')
                label = fluid.data(name='label', shape=[None, 1], dtype='int64')
1654 1655 1656
                reader = fluid.io.PyReader(feed_list=[image, label], capacity=4, iterable=True)

                user_defined_generator = random_image_and_label_generator(784, 784)
G
guofei 已提交
1657 1658 1659 1660 1661
                reader.decorate_batch_generator(user_defined_generator, fluid.CPUPlace())
                
                loss = network(image, label)
                executor = fluid.Executor(fluid.CPUPlace())
                executor.run(fluid.default_startup_program())
1662 1663 1664

                for _ in range(EPOCH_NUM):
                    for data in reader():
G
guofei 已提交
1665
                        executor.run(feed=data, fetch_list=[loss])
1666

S
add doc  
sneaxiy 已提交
1667
        '''
Z
Zeng Jinle 已提交
1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712
        self._loader.set_batch_generator(reader, places)


class DatasetLoader(DataLoaderBase):
    def __init__(self, dataset, places, drop_last):
        assert isinstance(dataset,
                          DatasetBase), "dataset must be type of DatasetBase"
        assert not in_dygraph_mode(
        ), "DatasetLoader is not supported in dygraph mode yet"

        thread_num = len(places)

        assert len(dataset.filelist) >= thread_num, \
            "Filelist number of dataset {} must be not less than place number {}".format(len(dataset.filelist), thread_num)

        if dataset.thread_num != 0 and dataset.thread_num != thread_num:
            logging.warn('thread_num {} which is set in Dataset is ignored'.
                         format(dataset.thread_num))

        dataset.set_thread(thread_num)

        if isinstance(dataset,
                      InMemoryDataset) and dataset.queue_num > thread_num:
            logging.warn("queue_num {} which is set in Dataset is ignored".
                         format(dataset.queue_num))
            dataset.set_queue_num(thread_num)

        self._dataset = dataset
        use_slots = [
            slot.name for slot in dataset.proto_desc.multi_slot_desc.slots
            if slot.is_used
        ]

        self._iterable_dataset = core.IterableDatasetWrapper(
            dataset.dataset, use_slots,
            _convert_places(places), dataset.proto_desc.batch_size, drop_last)

    def __iter__(self):
        self._dataset._finish_to_run()
        self._dataset._prepare_to_run()
        self._iterable_dataset._start()
        return self

    def __next__(self):
        return self._iterable_dataset._next()