test_dataloader.py

import math
import sys
import errno
import os
import ctypes
import torch
import gc
import time
import signal
import unittest
import itertools
import warnings
from torch import multiprocessing as mp
from torch.utils.data import _utils, Dataset, IterableDataset, TensorDataset, DataLoader, ConcatDataset, ChainDataset
from torch.utils.data._utils import MP_STATUS_CHECK_INTERVAL
from torch.utils.data.dataset import random_split
from torch._utils import ExceptionWrapper
from common_utils import (TestCase, run_tests, TEST_NUMPY, IS_WINDOWS, PY3,
                          IS_PYTORCH_CI, NO_MULTIPROCESSING_SPAWN, skipIfRocm,
                          load_tests, TEST_WITH_TSAN)

try:
    import psutil
    HAS_PSUTIL = True
except ImportError:
    HAS_PSUTIL = False
    err_msg = ("psutil not found. Some critical data loader tests relying on it "
               "(e.g., TestDataLoader.test_proper_exit) will not run.")
    if IS_PYTORCH_CI:
        raise ImportError(err_msg)
    else:
        warnings.warn(err_msg)

try:
    import faulthandler
    HAS_FAULTHANDLER = True
except ImportError:
    HAS_FAULTHANDLER = False
    err_msg = ("faulthandler not found. Some data loader tests use it for error "
               "reporting (e.g., TestDataLoader.test_proper_exit).")
    if IS_PYTORCH_CI:
        raise ImportError(err_msg)
    else:
        warnings.warn(err_msg)


# load_tests from common_utils is used to automatically filter tests for
# sharding on sandcastle. This line silences flake warnings
load_tests = load_tests

# We cannot import TEST_CUDA from common_cuda here, because if we do that,
# the TEST_CUDNN line from common_cuda will be executed multiple times
# as well during the execution of this test suite, and it will cause
# CUDA OOM error on Windows.
TEST_CUDA = torch.cuda.is_available()


if not NO_MULTIPROCESSING_SPAWN:
    # We want to use `spawn` if able because some of our tests check that the
    # data loader terminiates gracefully. To prevent hanging in the testing
    # process, such data loaders are run in a separate subprocess.
    #
    # We also want to test the `pin_memory=True` configuration, thus `spawn` is
    # required to launch such processes and they initialize the CUDA context.
    #
    # Mixing different start method is a recipe for disaster (e.g., using a fork
    # `mp.Event` with a spawn `mp.Process` segfaults). So we set this globally
    # to avoid bugs.
    #
    # Get a multiprocessing context because some test / third party library will
    # set start_method when imported, and setting again triggers `RuntimeError`.
    mp = mp.get_context(method='spawn')


# 60s of timeout?
# Yes, in environments where physical CPU resources are shared, e.g., CI, the
# time for a inter-process communication can be highly varying.  With 15~17s of
# timeout, we have observed flakiness in some CI builds (see
# pytorch/pytorch#14501, pytorch/pytorch#16608).  We follow the CPython
# multiprocessing setup and set the timeout to 60s here:
#
# https://github.com/python/cpython/blob/e8113f51a8bdf33188ee30a1c038a298329e7bfa/Lib/test/_test_multiprocessing.py#L73
JOIN_TIMEOUT = 60.0  # seconds


supported_multiprocessing_contexts = [None]
if torch.multiprocessing._supports_context:
    supported_multiprocessing_contexts += list(torch.multiprocessing.get_all_start_methods())


@unittest.skipIf(
    TEST_WITH_TSAN,
    "Fails with TSAN with the following error: starting new threads after multi-threaded "
    "fork is not supported. Dying (set die_after_fork=0 to override)")
class TestDatasetRandomSplit(TestCase):
    def test_lengths_must_equal_dataset_size(self):
        with self.assertRaises(ValueError):
            random_split([1, 2, 3, 4], [1, 2])

    def test_splits_have_correct_size(self):
        splits = random_split([1, 2, 3, 4, 5, 6], [2, 4])
        self.assertEqual(len(splits), 2)
        self.assertEqual(len(splits[0]), 2)
        self.assertEqual(len(splits[1]), 4)

    def test_splits_are_mutually_exclusive(self):
        data = [5, 2, 3, 4, 1, 6]
        splits = random_split(data, [2, 4])
        all_values = []
        all_values.extend(list(splits[0]))
        all_values.extend(list(splits[1]))
        data.sort()
        all_values.sort()
        self.assertListEqual(data, all_values)

    def test_splits_indexing_type(self):
        r"""Indices generated by random_split
          should be of integer type
        """
        class CustomDataset():
            def __init__(self, test_object, custom_list):
                self.data = custom_list
                self.test_object = test_object

            def __getitem__(self, key):
                self.test_object.assertEqual(type(key), type(0))
                return self.data[key]

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

        x = [1, 2, 3, 4, 5]
        dataset = CustomDataset(self, x)
        dataset = random_split(dataset, [5])[0]
        data_loader = DataLoader(dataset)
        for batch in data_loader:
            pass


class CUDACountingDataset(Dataset):
    def __init__(self, n):
        super(CUDACountingDataset, self).__init__()
        self.n = n

    def __getitem__(self, i):
        return torch.as_tensor(i, device='cuda')

    def __len__(self):
        return self.n


class CountingDataset(Dataset):
    def __init__(self, n):
        super(CountingDataset, self).__init__()
        self.n = n

    def __getitem__(self, i):
        return i

    def __len__(self):
        return self.n


class CountingIterableDataset(IterableDataset):
    def __init__(self, n):
        super(CountingIterableDataset, self).__init__()
        self.n = n

    def __iter__(self):
        return iter(range(self.n))

    def __len__(self):
        return self.n


@unittest.skipIf(
    TEST_WITH_TSAN,
    "Fails with TSAN with the following error: starting new threads after multi-threaded "
    "fork is not supported. Dying (set die_after_fork=0 to override)")
class TestTensorDataset(TestCase):

    def test_len(self):
        source = TensorDataset(torch.randn(15, 10, 2, 3, 4, 5), torch.randperm(15))
        self.assertEqual(len(source), 15)

    def test_getitem(self):
        t = torch.randn(15, 10, 2, 3, 4, 5)
        l = torch.randn(15, 10)
        source = TensorDataset(t, l)
        for i in range(15):
            self.assertEqual(t[i], source[i][0])
            self.assertEqual(l[i], source[i][1])

    def test_getitem_1d(self):
        t = torch.randn(15)
        l = torch.randn(15)
        source = TensorDataset(t, l)
        for i in range(15):
            self.assertEqual(t[i], source[i][0])
            self.assertEqual(l[i], source[i][1])

    def test_single_tensor(self):
        t = torch.randn(5, 10)
        source = TensorDataset(t)
        self.assertEqual(len(source), 5)
        for i in range(5):
            self.assertEqual(t[i], source[i][0])

    def test_many_tensors(self):
        t0 = torch.randn(5, 10, 2, 3, 4, 5)
        t1 = torch.randn(5, 10)
        t2 = torch.randn(5, 10, 2, 5)
        t3 = torch.randn(5, 10, 3, 7)
        source = TensorDataset(t0, t1, t2, t3)
        self.assertEqual(len(source), 5)
        for i in range(5):
            self.assertEqual(t0[i], source[i][0])
            self.assertEqual(t1[i], source[i][1])
            self.assertEqual(t2[i], source[i][2])
            self.assertEqual(t3[i], source[i][3])


@unittest.skipIf(
    TEST_WITH_TSAN,
    "Fails with TSAN with the following error: starting new threads after multi-threaded "
    "fork is not supported. Dying (set die_after_fork=0 to override)")
class TestConcatDataset(TestCase):

    def test_concat_two_singletons(self):
        result = ConcatDataset([[0], [1]])
        self.assertEqual(2, len(result))
        self.assertEqual(0, result[0])
        self.assertEqual(1, result[1])

    def test_concat_two_non_singletons(self):
        result = ConcatDataset([[0, 1, 2, 3, 4],
                                [5, 6, 7, 8, 9]])
        self.assertEqual(10, len(result))
        self.assertEqual(0, result[0])
        self.assertEqual(5, result[5])

    def test_concat_two_non_singletons_with_empty(self):
        # Adding an empty dataset somewhere is correctly handled
        result = ConcatDataset([[0, 1, 2, 3, 4],
                                [],
                                [5, 6, 7, 8, 9]])
        self.assertEqual(10, len(result))
        self.assertEqual(0, result[0])
        self.assertEqual(5, result[5])

    def test_concat_raises_index_error(self):
        result = ConcatDataset([[0, 1, 2, 3, 4],
                                [5, 6, 7, 8, 9]])
        with self.assertRaises(IndexError):
            # this one goes to 11
            result[11]

    def test_add_dataset(self):
        d1 = TensorDataset(torch.rand(7, 3, 28, 28), torch.rand(7))
        d2 = TensorDataset(torch.rand(7, 3, 28, 28), torch.rand(7))
        d3 = TensorDataset(torch.rand(7, 3, 28, 28), torch.rand(7))
        result = d1 + d2 + d3
        self.assertEqual(21, len(result))
        self.assertEqual(0, (d1[0][0] - result[0][0]).abs().sum())
        self.assertEqual(0, (d2[0][0] - result[7][0]).abs().sum())
        self.assertEqual(0, (d3[0][0] - result[14][0]).abs().sum())

    def test_iterable_dataset_err(self):
        d1 = TensorDataset(torch.rand(7, 3, 28, 28), torch.rand(7))
        it1 = CountingIterableDataset(5)
        it2 = CountingIterableDataset(10)

        with self.assertRaisesRegex(AssertionError, "does not support IterableDataset"):
            ConcatDataset([d1, it2, it1])

        with self.assertRaisesRegex(AssertionError, "does not support IterableDataset"):
            ConcatDataset([it2])

        with self.assertRaisesRegex(AssertionError, "does not support IterableDataset"):
            ConcatDataset([it1, d1])


# takes in dummy var so this can also be used as a `worker_init_fn`
def set_faulthander_if_available(_=None):
    if HAS_FAULTHANDLER:
        faulthandler.enable()
        if not IS_WINDOWS:
            # windows does not have faulthandler.register
            # chain=False prevents the default behavior of killing the process
            faulthandler.register(signal.SIGUSR1, chain=False)


set_faulthander_if_available()

# Process `pid` must have called `set_faulthander_if_available`
def print_traces_of_all_threads(pid):
    if HAS_FAULTHANDLER:
        if not IS_WINDOWS:
            # use the custom signal if available
            os.kill(pid, signal.SIGUSR1)
        else:
            # otherwise we can still use the handler given by faulthandler.enable()
            # at the cost of killing the process.
            os.kill(pid, signal.SIGSEGV)
    else:
        # if there is no faulthandler, use SIGINT otherwise and hope for the best
        os.kill(pid, signal.SIGINT)
    # wait in parent process to give subprocess some time to print
    time.sleep(5)


# The following `ErrorTrackingProcess` stores the first encountered exception in
# its `.exception` attribute.
# Inspired by https://stackoverflow.com/a/33599967
class ErrorTrackingProcess(mp.Process):

    # Why no *args?
    #   py2 doesn't support def fn(x, *args, key=val, **kwargs)
    # Setting disable_stderr=True may generate a lot of unrelated error outputs
    # but could be helpful for debugging.
    def __init__(self, disable_stderr=True, **kwargs):
        super(ErrorTrackingProcess, self).__init__(**kwargs)
        self._pconn, self._cconn = mp.Pipe()
        self._exception = None
        self.disable_stderr = disable_stderr

    def run(self):
        set_faulthander_if_available()
        if self.disable_stderr:
            # Disable polluting stderr with errors that are supposed to happen.
            sys.stderr = open(os.devnull, "w")
        try:
            super(ErrorTrackingProcess, self).run()
            self._cconn.send(None)
        except Exception:
            self._cconn.send(ExceptionWrapper(sys.exc_info()))
            raise

    def print_traces_of_all_threads(self):
        assert self.is_alive(), "can only use print_traces_of_all_threads if the process is alive"
        assert not self.disable_stderr, "do not disable stderr if you use print_traces_of_all_threads"
        # On platforms without `SIGUSR1`, `set_faulthander_if_available` sets
        # `faulthandler.enable()`, and `print_traces_of_all_threads` may kill
        # the process. So let's poll the exception first
        _ = self.exception
        print_traces_of_all_threads(self.pid)

    @property
    def exception(self):
        if self._pconn.poll():
            self._exception = self._pconn.recv()
        if self._exception is None:
            return None
        else:
            return self._exception.exc_type(self._exception.exc_msg)

    # ESRCH means that os.kill can't finds alive proc
    def send_signal(self, signum, ignore_ESRCH=False):
        try:
            os.kill(self.pid, signum)
        except OSError as e:
            if not ignore_ESRCH or e.errno != errno.ESRCH:
                raise


class ErrorDataset(Dataset):

    def __init__(self, size):
        self.size = size

    def __len__(self):
        return self.size


class SegfaultDataset(Dataset):

    def __init__(self, size):
        self.size = size

    def __getitem__(self, idx):
        return ctypes.string_at(0)

    def __len__(self):
        return self.size


class SleepDataset(Dataset):

    def __init__(self, size, sleep_sec):
        self.size = size
        self.sleep_sec = sleep_sec
        self.sleeped = False

    def __getitem__(self, idx):
        if not self.sleeped:
            time.sleep(self.sleep_sec)
            self.sleeped = True
        return idx

    def __len__(self):
        return self.size


class SeedDataset(Dataset):

    def __init__(self, size):
        self.size = size

    def __getitem__(self, idx):
        return torch.initial_seed()

    def __len__(self):
        return self.size


class WorkerSpecificIterableDataset(IterableDataset):
    def __init__(self, sizes_for_all_workers):
        self.sizes_for_all_workers = sizes_for_all_workers

    def __iter__(self):
        worker_info = torch.utils.data.get_worker_info()
        assert worker_info is not None
        return iter(range(self.sizes_for_all_workers[worker_info.id]))


# Inspired by https://stackoverflow.com/a/26703365
# If all workers will call `sync_once`, they will be blocked until all workers
# reach the call (i.e., acting like a barrier).
# This can be used to ensure that each worker at least processes one data.
class SynchronizedDataset(Dataset):

    def __init__(self, size, batch_size, num_workers):
        assert size >= num_workers * batch_size
        self.count = mp.Value('i', 0, lock=True)
        self.barrier = mp.Semaphore(0)
        self.num_workers = num_workers
        self.size = size

    def sync_once(self):
        with self.count.get_lock():
            self.count.value += 1
            if self.count.value == self.num_workers:
                self.barrier.release()
        self.barrier.acquire()
        self.barrier.release()

    def __getitem__(self, idx):
        raise NotImplementedError

    def __len__(self):
        return self.size


class SynchronizedSeedDataset(SynchronizedDataset):
    def __getitem__(self, idx):
        self.sync_once()
        return torch.initial_seed()


def _test_timeout():
    dataset = SleepDataset(10, 3)
    dataloader = DataLoader(dataset, batch_size=2, num_workers=2, timeout=1)
    _ = next(iter(dataloader))


def _test_timeout_pin_memory():
    dataset = SleepDataset(10, 3)
    dataloader = DataLoader(dataset, batch_size=2, num_workers=2, timeout=1, pin_memory=True)
    _ = next(iter(dataloader))


def disable_stderr(worker_id):
    r"""
    Avoids printing "ERROR: Unexpected segmentation fault encountered in worker."
    from workers. Since worker signal handler prints with low-level write(),
    this has to be done on OS level via dup.

    This is used as worker_init_fn for test_segfault.
    """
    sys.stderr.flush()  # flush library buffers that dup2 knows nothing about
    # Can't use a with-block because otherwise the fd will be closed when this
    # function ends.
    devnull = open(os.devnull, 'w')
    os.dup2(devnull.fileno(), sys.stderr.fileno())


def _test_segfault():
    dataset = SegfaultDataset(10)
    dataloader = DataLoader(dataset, batch_size=2, num_workers=2, worker_init_fn=disable_stderr)
    _ = next(iter(dataloader))


class TestProperExitDataset(Dataset):
    def __init__(self, size, error_event):
        self.size = size
        self.error_event = error_event

    def __len__(self):
        return self.size

    def __getitem__(self, idx):
        worker_info = torch.utils.data.get_worker_info()
        if self.error_event is not None and self.error_event.is_set() and \
                worker_info.id == worker_info.num_workers - 1:
            # only error in the last worker
            raise RuntimeError('Worker error')
        return torch.tensor([idx])


class TestProperExitIterableDataset(IterableDataset):
    def __init__(self, size, error_event):
        self.error_event = error_event
        self.size = size
        self.remaining = size

    def __len__(self):
        return self.size

    def __iter__(self):
        return self

    def __next__(self):
        worker_info = torch.utils.data.get_worker_info()
        if self.error_event is not None and self.error_event.is_set() and \
                worker_info.id == worker_info.num_workers - 1:
            # only error in the last worker
            raise RuntimeError('Worker error')
        self.remaining -= 1
        if self.remaining < 0:
            raise StopIteration
        return torch.tensor(-1000)

    next = __next__  # py2 compatibility


# See TestDataLoader.test_proper_exit for usage
def _test_proper_exit(is_iterable_dataset, use_workers, pin_memory, exit_method,
                      hold_iter_reference, loader_setup_event, tester_setup_event):
    num_workers = 2 if use_workers else 0

    if exit_method == 'worker_error' or exit_method == 'worker_kill':
        assert use_workers is True

    if exit_method == 'worker_error':
        worker_error_event = mp.Event()
    else:
        worker_error_event = None

    if is_iterable_dataset:
        ds = TestProperExitIterableDataset(7, worker_error_event)
    else:
        ds = TestProperExitDataset(12, worker_error_event)

    loader = DataLoader(ds, batch_size=1, shuffle=False,
                        num_workers=num_workers, pin_memory=pin_memory,
                        worker_init_fn=set_faulthander_if_available)

    error_it = 2

    if use_workers:
        # 2 is the magical per-worker prefetch number...
        # FIXME: change this after the number becomes configurable.
        if is_iterable_dataset:
            assert len(ds) * num_workers > (error_it + 2 + 1)
        else:
            assert len(loader) > (error_it + 2 + 1) * num_workers
    else:
        if is_iterable_dataset:
            assert len(ds) > error_it + 1
        else:
            assert len(loader) > error_it + 1

    it = iter(loader)
    if use_workers:
        workers = it._workers

    def kill_pid(pid):
        psutil_p = psutil.Process(pid)
        psutil_p.kill()
        psutil_p.wait(JOIN_TIMEOUT)
        assert not psutil_p.is_running()

    for i, _ in enumerate(it):
        if i == 0:
            if not hold_iter_reference:
                del it
            loader_setup_event.set()
            tester_setup_event.wait()
            # ensure that the workers are still alive
            if use_workers:
                for w in workers:
                    assert w.is_alive()
            if worker_error_event is not None:
                worker_error_event.set()

        if i == error_it:
            if exit_method == 'loader_error':
                raise RuntimeError('Loader error')
            elif exit_method == 'loader_kill':
                kill_pid(os.getpid())
            elif exit_method == 'worker_kill':
                kill_pid(workers[-1].pid)  # kill last worker

    if not hold_iter_reference:
        # Tries to trigger the __del__ clean-up rather than the automatic
        # exiting of daemonic children. Technically it should be automatically
        # triggered, but I don't want to rely on the implementation detail of
        # Python gc.
        gc.collect()


class TestWorkerInfoDataset(SynchronizedDataset):
    def __getitem__(self, idx):
        self.sync_once()
        return torch.tensor(self.value)


# Should be used as worker_init_fn with TestWorkerInfoDataset.
# See _test_get_worker_info below for usage.
def test_worker_info_init_fn(worker_id):
    worker_info = torch.utils.data.get_worker_info()
    assert worker_id == worker_info.id, "worker_init_fn and worker_info should have consistent id"
    assert worker_id < worker_info.num_workers, "worker_init_fn and worker_info should have valid id"
    assert worker_info.seed == torch.initial_seed(), "worker_init_fn and worker_info should have consistent seed"
    dataset = worker_info.dataset
    assert isinstance(dataset, TestWorkerInfoDataset), "worker_info should have correct dataset copy"
    assert not hasattr(dataset, 'value'), "worker_info should have correct dataset copy"
    # test that WorkerInfo attributes are read-only
    try:
        worker_info.id = 3999
    except RuntimeError as e:
        assert str(e) == "Cannot assign attributes to WorkerInfo objects"
    try:
        worker_info.a = 3
    except RuntimeError as e:
        assert str(e) == "Cannot assign attributes to WorkerInfo objects"
    dataset.value = [worker_id, os.getpid()]


def _test_get_worker_info():
    # get_worker_info returns None in main proc
    assert torch.utils.data.get_worker_info() is None
    num_workers = 2
    batch_size = 2
    dataset = TestWorkerInfoDataset(6, batch_size, num_workers)
    dataloader = DataLoader(dataset, batch_size=batch_size,
                            num_workers=num_workers,
                            worker_init_fn=test_worker_info_init_fn)
    it = iter(dataloader)
    data = []
    for d in it:
        data.append(d)
    worker_pids = [w.pid for w in it._workers]
    data = torch.cat(data, 0)
    for d in data:
        # each `d` is a [worker_id, worker_pid] pair, which is set in
        # test_worker_info_init_fn
        assert d[1] == worker_pids[d[0]]
    # get_worker_info returns None in main proc after data loading
    assert torch.utils.data.get_worker_info() is None
    # main proc dataset was never assigned this attribute
    assert not hasattr(dataset, 'value')
    try:
        _ = dataset[0]
    except AttributeError:
        return
    raise RuntimeError('Expected AttributeError')


# test custom init function
def init_fn(worker_id):
    torch.manual_seed(12345)


# used with test_error_in_init
class ErrorIterableDataset(IterableDataset):
    def __iter__(self):
        raise RuntimeError("Error in __iter__")


# used with test_error_in_init
def error_worker_init_fn(_):
    raise RuntimeError("Error in worker_init_fn")


@unittest.skipIf(
    TEST_WITH_TSAN,
    "Fails with TSAN with the following error: starting new threads after multi-threaded "
    "fork is not supported. Dying (set die_after_fork=0 to override)")
class TestDataLoader(TestCase):

    def setUp(self):
        super(TestDataLoader, self).setUp()
        self.data = torch.randn(100, 2, 3, 5)
        self.labels = torch.randperm(50).repeat(2)
        self.dataset = TensorDataset(self.data, self.labels)

    def _test_sequential(self, loader):
        batch_size = loader.batch_size
        if batch_size is None:
            for idx, (sample, target) in enumerate(loader):
                self.assertEqual(sample, self.data[idx])
                self.assertEqual(target, self.labels[idx])
            self.assertEqual(idx, len(self.dataset) - 1)
        else:
            for i, (sample, target) in enumerate(loader):
                idx = i * batch_size
                self.assertEqual(sample, self.data[idx:idx + batch_size])
                self.assertEqual(target, self.labels[idx:idx + batch_size])
            self.assertEqual(i, math.floor((len(self.dataset) - 1) / batch_size))

    def _test_shuffle(self, loader):
        found_data = {i: 0 for i in range(self.data.size(0))}
        found_labels = {i: 0 for i in range(self.labels.size(0))}
        batch_size = loader.batch_size
        for i, (batch_samples, batch_targets) in enumerate(loader):
            for sample, target in zip(batch_samples, batch_targets):
                for data_point_idx, data_point in enumerate(self.data):
                    if data_point.eq(sample).all():
                        self.assertFalse(found_data[data_point_idx])
                        found_data[data_point_idx] += 1
                        break
                self.assertEqual(target, self.labels[data_point_idx])
                found_labels[data_point_idx] += 1
            self.assertEqual(sum(found_data.values()), (i + 1) * batch_size)
            self.assertEqual(sum(found_labels.values()), (i + 1) * batch_size)
        self.assertEqual(i, math.floor((len(self.dataset) - 1) / batch_size))

    def _test_error(self, loader):
        it = iter(loader)
        errors = 0
        while True:
            try:
                next(it)
            except NotImplementedError:
                errors += 1
            except StopIteration:
                self.assertEqual(errors,
                                 math.ceil(float(len(loader.dataset)) / loader.batch_size))
                return

    def test_error_in_init(self):
        for num_workers in [0, 2]:
            loader = DataLoader(ErrorIterableDataset(), num_workers=num_workers)
            with self.assertRaisesRegex(RuntimeError, 'Error in __iter__'):
                list(iter(loader))

        loader = DataLoader(self.dataset, num_workers=2, worker_init_fn=error_worker_init_fn)
        with self.assertRaisesRegex(RuntimeError, 'Error in worker_init_fn'):
            list(iter(loader))

    def test_invalid_assign_after_init(self):
        dl = DataLoader(self.dataset)
        for attr in ('batch_size', 'sampler', 'batch_sampler', 'drop_last', 'dataset'):
            def fn():
                setattr(dl, attr, {})

            self.assertRaises(ValueError, fn)

    def test_sequential_nonbatch(self):
        self._test_sequential(DataLoader(self.dataset, batch_size=None))

    def test_sequential_batch(self):
        self._test_sequential(DataLoader(self.dataset))
        self._test_sequential(DataLoader(self.dataset, batch_size=2))

    def test_growing_dataset(self):
        dataset = [torch.ones(4) for _ in range(4)]
        dataloader_seq = DataLoader(dataset, shuffle=False)
        dataloader_shuffle = DataLoader(dataset, shuffle=True)
        dataset.append(torch.ones(4))
        self.assertEqual(len(dataloader_seq), 5)
        self.assertEqual(len(dataloader_shuffle), 5)

    @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
    def test_sequential_pin_memory(self):
        loader = DataLoader(self.dataset, batch_size=2, pin_memory=True)
        for input, target in loader:
            self.assertTrue(input.is_pinned())
            self.assertTrue(target.is_pinned())

    def test_multiple_dataloaders(self):
        for multiprocessing_context in supported_multiprocessing_contexts:
            loader1_it = iter(DataLoader(self.dataset, num_workers=1))
            loader2_it = iter(DataLoader(self.dataset, num_workers=2, multiprocessing_context=multiprocessing_context))
            next(loader1_it)
            next(loader1_it)
            next(loader2_it)
            next(loader2_it)
            next(loader1_it)
            next(loader2_it)

    @unittest.skip("temporarily disable until flaky failures are fixed")
    def test_segfault(self):
        p = ErrorTrackingProcess(target=_test_segfault)
        p.start()
        p.join(JOIN_TIMEOUT)
        try:
            self.assertFalse(p.is_alive())
            self.assertNotEqual(p.exitcode, 0)
            if IS_WINDOWS:
                self.assertIsInstance(p.exception, OSError)
                self.assertRegex(str(p.exception), r'access violation reading ')
            else:
                self.assertIsInstance(p.exception, RuntimeError)
                self.assertRegex(str(p.exception), r'DataLoader worker \(pid \d+\) is killed by signal: ')
        finally:
            p.terminate()

    def test_timeout(self):
        if TEST_CUDA and not NO_MULTIPROCESSING_SPAWN:
            # This test runs in a subprocess, which can only initialize CUDA with spawn.
            # _test_timeout_pin_memory with pin_memory=True initializes CUDA when the iterator is
            # constructed.
            targets = (_test_timeout, _test_timeout_pin_memory)
        else:
            targets = (_test_timeout,)
        for target in targets:
            p = ErrorTrackingProcess(target=target)
            p.start()
            p.join(JOIN_TIMEOUT)
            try:
                self.assertFalse(p.is_alive())
                self.assertNotEqual(p.exitcode, 0)
                self.assertIsInstance(p.exception, RuntimeError)
                self.assertRegex(str(p.exception), r'DataLoader timed out after \d+ seconds')
            finally:
                p.terminate()

    def test_invalid_ctor_args_combinations(self):
        # general
        with self.assertRaisesRegex(ValueError, "num_workers option should be non-negative"):
            DataLoader(self.dataset, num_workers=-1)
        with self.assertRaisesRegex(ValueError, "timeout option should be non-negative"):
            DataLoader(self.dataset, timeout=-1)

        if torch.multiprocessing._supports_context:
            valid_ctx = list(torch.multiprocessing.get_all_start_methods())[-1]
            with self.assertRaisesRegex(ValueError, r"multi-process loading \(num_workers > 0\), but got"):
                DataLoader(self.dataset, num_workers=0, multiprocessing_context=valid_ctx)
            with self.assertRaisesRegex(ValueError, "should specify a valid start method in"):
                DataLoader(self.dataset, num_workers=1, multiprocessing_context='bad')
            with self.assertRaisesRegex(ValueError, "multiprocessing_context option should be a valid context "):
                DataLoader(self.dataset, num_workers=1, multiprocessing_context=object())
        else:
            with self.assertRaisesRegex(ValueError, "multiprocessing_context relies on Python >= 3.4"):
                DataLoader(self.dataset, num_workers=1, multiprocessing_context='fork')

        # map-style
        sampler = torch.utils.data.SequentialSampler(self.dataset)
        batch_sampler = torch.utils.data.BatchSampler(sampler, 3, False)
        with self.assertRaisesRegex(ValueError, "sampler option is mutually exclusive with shuffle"):
            DataLoader(self.dataset, batch_size=11, sampler=sampler, shuffle=True)
        with self.assertRaisesRegex(ValueError, "sampler option is mutually exclusive with shuffle"):
            DataLoader(self.dataset, batch_sampler=batch_sampler, sampler=sampler, shuffle=True)
        with self.assertRaisesRegex(ValueError, "sampler option is mutually exclusive with shuffle"):
            DataLoader(self.dataset, batch_sampler=batch_sampler, sampler=sampler, shuffle=3)
        with self.assertRaisesRegex(ValueError, "batch_sampler option is mutually exclusive with"):
            DataLoader(self.dataset, batch_size=11, batch_sampler=batch_sampler)
        with self.assertRaisesRegex(ValueError, "batch_sampler option is mutually exclusive with"):
            DataLoader(self.dataset, shuffle=True, batch_sampler=batch_sampler)
        with self.assertRaisesRegex(ValueError, "batch_sampler option is mutually exclusive with"):
            DataLoader(self.dataset, drop_last=True, batch_sampler=batch_sampler)
        with self.assertRaisesRegex(ValueError, "batch_sampler option is mutually exclusive with"):
            DataLoader(self.dataset, drop_last=3, batch_sampler=batch_sampler)

        # iterable-style
        dataset = CountingIterableDataset(20)
        with self.assertRaisesRegex(ValueError, "DataLoader with IterableDataset: expected unspecified shuffle"):
            DataLoader(dataset, shuffle=True)
        with self.assertRaisesRegex(ValueError, "DataLoader with IterableDataset: expected unspecified shuffle"):
            DataLoader(dataset, shuffle=3)
        with self.assertRaisesRegex(ValueError, "DataLoader with IterableDataset: expected unspecified sampler"):
            DataLoader(dataset, sampler=torch.utils.data.SequentialSampler(dataset))
        with self.assertRaisesRegex(ValueError, "DataLoader with IterableDataset: expected unspecified sampler"):
            DataLoader(dataset, sampler=3)
        with self.assertRaisesRegex(ValueError, "DataLoader with IterableDataset: expected unspecified batch_sampler"):
            DataLoader(dataset, batch_sampler=torch.utils.data.BatchSampler(
                torch.utils.data.SequentialSampler(dataset), 3, False))
        with self.assertRaisesRegex(ValueError, "DataLoader with IterableDataset: expected unspecified batch_sampler"):
            DataLoader(dataset, batch_sampler=3)

    def test_builtin_collection_conversion(self):
        for coll_ty in (list, tuple):
            for num_workers in (0, 1):
                # map-style dataset
                dataset = CountingDataset(20)
                # no auto-batching
                fetched = coll_ty(DataLoader(dataset, batch_size=None, num_workers=num_workers))
                self.assertEqual(fetched, coll_ty(range(20)))
                # auto-batching
                fetched = coll_ty(DataLoader(dataset, batch_size=2, num_workers=num_workers))
                self.assertEqual(fetched, coll_ty(torch.tensor([i, i + 1]) for i in range(0, 20, 2)))

                # iterable-style dataset
                dataset = CountingIterableDataset(20)
                # no auto-batching
                fetched = coll_ty(DataLoader(dataset, batch_size=None, num_workers=num_workers))
                self.assertEqual(fetched, coll_ty(range(20)))
                # auto-batching
                # this IterableDataset isn't configured for each worker, so for
                # the equality test below to be valid, we cannot have more than 1 workers.
                assert num_workers in [0, 1], "invalid test"
                fetched = coll_ty(DataLoader(dataset, batch_size=2, num_workers=num_workers))
                self.assertEqual(fetched, coll_ty(torch.tensor([i, i + 1]) for i in range(0, 20, 2)))

    def test_iterable_style_dataset(self):
        # [no auto-batching] single process loading
        dataset = CountingIterableDataset(20)
        dataloader = DataLoader(dataset, batch_size=None)
        fetched = list(dataloader)
        self.assertEqual(len(fetched), 20)
        for i, d in enumerate(fetched):
            # non-batched should not convert ints into tensors
            self.assertIsInstance(d, torch._six.int_classes)
            self.assertEqual(d, i)
        with self.assertRaisesRegex(TypeError, "Cannot determine the DataLoader length of a IterableDataset"):
            len(dataloader)  # DataLoader with iterable-style dataset should error in __len__

        # [no auto-batching] multiprocessing loading
        num_workers = 3
        sizes_for_all_workers = [0, 4, 20]
        expected = sorted(sum((list(range(s)) for s in sizes_for_all_workers), []))
        assert len(sizes_for_all_workers) == num_workers, 'invalid test case'
        dataset = WorkerSpecificIterableDataset(sizes_for_all_workers)
        dataloader = DataLoader(dataset, num_workers=num_workers, batch_size=None,
                                worker_init_fn=set_faulthander_if_available)
        dataloader_iter = iter(dataloader)
        fetched = sorted([d for d in dataloader_iter])
        for a, b in zip(fetched, expected):
            # non-batched should not convert ints into tensors
            self.assertIsInstance(a, torch._six.int_classes)
            self.assertEqual(a, b)
        with self.assertRaisesRegex(TypeError, "Cannot determine the DataLoader length of a IterableDataset"):
            len(dataloader)  # DataLoader with iterable-style dataset should error in __len__

        # [no auto-batching] test that workers exit gracefully
        workers = dataloader_iter._workers
        del dataloader_iter
        try:
            for w in workers:
                w.join(JOIN_TIMEOUT)
                self.assertFalse(w.is_alive())
                self.assertEqual(w.exitcode, 0)
        finally:
            for w in workers:
                w.terminate()

        # [auto-batching] single process loading
        dataset = CountingIterableDataset(20)
        fetched = list(DataLoader(dataset, batch_size=7))
        self.assertEqual(len(fetched), 3)
        self.assertEqual(fetched[0].tolist(), list(range(7)))
        self.assertEqual(fetched[1].tolist(), list(range(7, 14)))
        self.assertEqual(fetched[2].tolist(), list(range(14, 20)))

        # [auto-batching] multiprocessing loading
        num_workers = 3
        sizes_for_all_workers = [0, 4, 20]
        expected = sorted(sum((list(range(s)) for s in sizes_for_all_workers), []))
        assert len(sizes_for_all_workers) == num_workers, 'invalid test case'
        dataset = WorkerSpecificIterableDataset(sizes_for_all_workers)
        # worker 0 should return 0 batches
        # worker 1 should return 1 batches
        # worker 2 should return 3 batches
        dataloader = DataLoader(dataset, num_workers=num_workers, batch_size=7)
        dataloader_iter = iter(dataloader)
        fetched = list(dataloader_iter)
        self.assertEqual(len(fetched), 4)
        fetched = set(tuple(t.tolist()) for t in fetched)
        self.assertEqual(fetched, {tuple(range(4)), tuple(range(7)), tuple(range(7, 14)), tuple(range(14, 20))})

        # [auto-batching] test that workers exit gracefully
        workers = dataloader_iter._workers
        del dataloader_iter
        try:
            for w in workers:
                w.join(JOIN_TIMEOUT)
                self.assertFalse(w.is_alive())
                self.assertEqual(w.exitcode, 0)
        finally:
            for w in workers:
                w.terminate()

        # [auto-batching & drop_last] single process loading
        dataset = CountingIterableDataset(20)
        fetched = list(DataLoader(dataset, batch_size=7, drop_last=True))
        self.assertEqual(len(fetched), 2)
        self.assertEqual(fetched[0].tolist(), list(range(7)))
        self.assertEqual(fetched[1].tolist(), list(range(7, 14)))

        # [auto-batching & drop_last] multiprocessing loading
        num_workers = 3
        sizes_for_all_workers = [0, 4, 20]
        expected = sorted(sum((list(range(s)) for s in sizes_for_all_workers), []))
        assert len(sizes_for_all_workers) == num_workers, 'invalid test case'
        dataset = WorkerSpecificIterableDataset(sizes_for_all_workers)
        # worker 0 should return 0 batches
        # worker 1 should return 1 batches
        # worker 2 should return 3 batches
        dataloader = DataLoader(dataset, num_workers=num_workers, batch_size=7, drop_last=True,
                                worker_init_fn=set_faulthander_if_available)
        dataloader_iter = iter(dataloader)
        fetched = list(dataloader_iter)
        self.assertEqual(len(fetched), 2)
        fetched = set(tuple(t.tolist()) for t in fetched)
        self.assertEqual(fetched, {tuple(range(7)), tuple(range(7, 14))})

        # [auto-batching & drop_last] test that workers exit gracefully
        workers = dataloader_iter._workers
        del dataloader_iter
        try:
            for w in workers:
                w.join(JOIN_TIMEOUT)
                self.assertFalse(w.is_alive())
                self.assertEqual(w.exitcode, 0)
        finally:
            for w in workers:
                w.terminate()

    def test_chain_iterable_style_dataset(self):
        # chaining (concatenation)
        dataset1 = CountingIterableDataset(20)
        dataset2 = CountingIterableDataset(15)
        expected = list(range(20)) + list(range(15))
        for num_workers in [0, 1]:
            for chained_dataset in [dataset1 + dataset2, ChainDataset([dataset1, dataset2])]:
                fetched = list(DataLoader(chained_dataset, num_workers=num_workers))
                self.assertEqual(len(fetched), len(expected))
                for e, d in zip(expected, fetched):
                    self.assertIsInstance(d, torch.Tensor)
                    self.assertEqual(e, d)

        with self.assertRaisesRegex(AssertionError, "ChainDataset only supports IterableDataset"):
            list(iter(dataset1 + self.dataset))

        with self.assertRaisesRegex(AssertionError, "ChainDataset only supports IterableDataset"):
            list(iter(ChainDataset([dataset1, self.dataset])))

    def test_multiprocessing_contexts(self):
        reference = [
            torch.arange(3),
            torch.arange(3, 6),
            torch.arange(6, 9),
            torch.arange(9, 11),
        ]
        counting_ds_n = 11
        dl_common_args = dict(num_workers=3, batch_size=3, pin_memory=(not TEST_CUDA))
        for ctx in supported_multiprocessing_contexts:
            if ctx in ['spawn', 'forkserver'] and TEST_CUDA and not IS_WINDOWS:  # windows doesn't support sharing cuda tensor
                dl_cls = CUDACountingDataset
            else:
                ds_cls = CountingDataset
            self.assertEqual(
                reference, list(DataLoader(ds_cls(counting_ds_n), multiprocessing_context=ctx, **dl_common_args)))
            if ctx is not None:
                # test ctx object
                ctx = mp.get_context(ctx)
                self.assertEqual(
                    reference, list(DataLoader(ds_cls(counting_ds_n), multiprocessing_context=ctx, **dl_common_args)))

    def test_worker_seed(self):
        num_workers = 6
        batch_size = 1
        dataset = SynchronizedSeedDataset(num_workers, batch_size, num_workers)
        dataloader = DataLoader(dataset, batch_size=batch_size, num_workers=num_workers)
        seeds = set()
        for batch in dataloader:
            seeds.add(batch[0])
        self.assertEqual(len(seeds), num_workers)

    def test_worker_init_fn(self):
        dataset = SeedDataset(4)
        dataloader = DataLoader(dataset, batch_size=2, num_workers=2,
                                worker_init_fn=init_fn)
        for batch in dataloader:
            self.assertEqual(12345, batch[0])
            self.assertEqual(12345, batch[1])

    def test_get_worker_info(self):
        p = ErrorTrackingProcess(target=_test_get_worker_info)
        p.start()
        p.join(JOIN_TIMEOUT)
        try:
            self.assertFalse(p.is_alive())
            self.assertEqual(p.exitcode, 0)
        finally:
            p.terminate()

    def test_shuffle(self):
        self._test_shuffle(DataLoader(self.dataset, shuffle=True))

    def test_shuffle_batch(self):
        self._test_shuffle(DataLoader(self.dataset, batch_size=2, shuffle=True))

    def test_sequential_workers(self):
        self._test_sequential(DataLoader(self.dataset, num_workers=4))

    def test_seqential_batch_workers(self):
        self._test_sequential(DataLoader(self.dataset, batch_size=2, num_workers=4))

    def test_shuffle_workers(self):
        self._test_shuffle(DataLoader(self.dataset, shuffle=True, num_workers=4))

    def test_shuffle_batch_workers(self):
        self._test_shuffle(DataLoader(self.dataset, batch_size=2, shuffle=True, num_workers=4))

    def test_RandomSampler(self):

        from collections import Counter
        from torch.utils.data import RandomSampler

        def sample_stat(sampler, num_samples):
            counts = Counter(sampler)
            count_repeated = sum(val > 1 for val in counts.values())
            return (count_repeated, min(counts.keys()), max(counts.keys()))

        # test sample with replacement
        n = len(self.dataset) + 1  # ensure at least one sample is drawn more than once
        sampler_with_replacement = RandomSampler(self.dataset, replacement=True, num_samples=n)
        count_repeated, minval, maxval = sample_stat(sampler_with_replacement, n)
        self.assertTrue(count_repeated > 0)
        self.assertTrue(minval >= 0)
        self.assertTrue(maxval < len(self.dataset))

        # test sample without replacement
        sampler_without_replacement = RandomSampler(self.dataset)
        count_repeated, minval, maxval = sample_stat(sampler_without_replacement, len(self.dataset))
        self.assertTrue(count_repeated == 0)
        self.assertTrue(minval == 0)
        self.assertTrue(maxval == len(self.dataset) - 1)

        # raise error when replacement=False and num_samples is not None
        self.assertRaises(ValueError, lambda: RandomSampler(self.dataset, num_samples=len(self.dataset)))

        self.assertRaises(ValueError, lambda: RandomSampler(self.dataset, num_samples=0))

    def test_random_sampler_len_with_replacement(self):
        from torch.utils.data import RandomSampler
        # add 5 extra samples
        num_samples = len(self.dataset) + 5
        sampler = RandomSampler(self.dataset,
                                replacement=True,
                                num_samples=num_samples)
        # test len method
        self.assertEqual(num_samples, len(sampler))

        # test with iteration
        count_num_samples = sum(1 for _ in sampler)
        self.assertEqual(num_samples, count_num_samples)

        # test with dataloader, batch_size = 1
        batch_size = 1
        count_num_samples_in_data_loader = len(DataLoader(
            self.dataset, batch_size=batch_size, sampler=sampler))
        self.assertEqual(num_samples, count_num_samples_in_data_loader)

        # test with dataloader, batch_size = 6
        batch_size = 6
        count_num_samples_in_data_loader = len(DataLoader(
            self.dataset, batch_size=batch_size, sampler=sampler))
        self.assertEqual(int(math.ceil(float(num_samples) / batch_size)),
                         count_num_samples_in_data_loader)

    def test_duplicating_data_with_drop_last(self):

        from torch.utils.data.distributed import DistributedSampler

        num_processes = 4
        num_batches = 9
        data_set = torch.IntTensor(range(num_batches))
        scanned_data = torch.IntTensor([])
        for i in range(num_processes):
            s = DistributedSampler(data_set, num_processes, i)
            d_loader = DataLoader(data_set, batch_size=int(num_batches / num_processes), drop_last=True, sampler=s)
            for data in d_loader:
                scanned_data = torch.cat((scanned_data, data), 0)

        self.assertEqual(scanned_data.size(), scanned_data.unique().size())

    def _test_batch_sampler(self, **kwargs):
        # [(0, 1), (2, 3, 4), (5, 6), (7, 8, 9), ...]
        batches = []
        for i in range(0, 20, 5):
            batches.append(tuple(range(i, i + 2)))
            batches.append(tuple(range(i + 2, i + 5)))

        dl = DataLoader(self.dataset, batch_sampler=batches, **kwargs)
        self.assertEqual(len(dl), 8)
        for i, (input, _target) in enumerate(dl):
            if i % 2 == 0:
                offset = i * 5 // 2
                self.assertEqual(len(input), 2)
                self.assertEqual(input, self.data[offset:offset + 2])
            else:
                offset = i * 5 // 2
                self.assertEqual(len(input), 3)
                self.assertEqual(input, self.data[offset:offset + 3])

    def test_batch_sampler(self):
        self._test_batch_sampler()
        self._test_batch_sampler(num_workers=4)
        if not NO_MULTIPROCESSING_SPAWN and torch.multiprocessing._supports_context:
            self._test_batch_sampler(num_workers=4, multiprocessing_context='spawn')

    @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
    def test_shuffle_pin_memory(self):
        loader = DataLoader(self.dataset, batch_size=2, shuffle=True, num_workers=4, pin_memory=True)
        for input, target in loader:
            self.assertTrue(input.is_pinned())
            self.assertTrue(target.is_pinned())

    @unittest.skipIf(not TEST_NUMPY, "numpy unavailable")
    def test_numpy(self):
        import numpy as np

        class TestDataset(torch.utils.data.Dataset):
            def __getitem__(self, i):
                return np.ones((2, 3, 4)) * i

            def __len__(self):
                return 1000

        loader = DataLoader(TestDataset(), batch_size=12)
        batch = next(iter(loader))
        self.assertIsInstance(batch, torch.DoubleTensor)
        self.assertEqual(batch.size(), torch.Size([12, 2, 3, 4]))

    def test_error(self):
        self._test_error(DataLoader(ErrorDataset(100), batch_size=2, shuffle=True))

    def test_error_workers(self):
        self._test_error(DataLoader(ErrorDataset(41), batch_size=2, shuffle=True, num_workers=4))

    @unittest.skipIf(IS_WINDOWS, "FIXME: stuck test")
    def test_partial_workers(self):
        r"""Check that workers exit even if the iterator is not exhausted."""
        if TEST_CUDA:
            pin_memory_configs = (True, False)
        else:
            pin_memory_configs = (False,)

        for pin_memory in pin_memory_configs:
            loader = iter(DataLoader(self.dataset, batch_size=2, num_workers=4, pin_memory=pin_memory))
            workers = loader._workers
            if pin_memory:
                pin_memory_thread = loader._pin_memory_thread
            for i, _ in enumerate(loader):
                if i == 10:
                    break
            assert i == 10
            del loader
            for w in workers:
                w.join(JOIN_TIMEOUT)
                self.assertFalse(w.is_alive(), 'subprocess not terminated')
            if pin_memory:
                pin_memory_thread.join(JOIN_TIMEOUT)
                self.assertFalse(pin_memory_thread.is_alive())

    @skipIfRocm
    @unittest.skipIf(not HAS_PSUTIL, "psutil not found")
    def test_proper_exit(self):
        (r'''There might be ConnectionResetError or leaked semaphore warning '''
         r'''(due to dirty process exit), but they are all safe to ignore''')

        # TODO: test the case where the pin_memory_thread triggers an
        #       error/fatal signal. I haven't found out how to properly do that.

        for is_iterable_dataset, use_workers, pin_memory, hold_iter_reference in \
                itertools.product([True, False], repeat=4):

            # `hold_iter_reference` specifies whether we hold a reference to the
            # iterator. This is interesting because Python3 error traces holds a
            # reference to the frames, which hold references to all the local
            # variables including the iterator, and then the iterator dtor may
            # not be called before process end. It is important to see that the
            # processes still exit in both cases.

            if pin_memory and (not TEST_CUDA or NO_MULTIPROCESSING_SPAWN or IS_WINDOWS):
                # This test runs in a subprocess, which can only initialize CUDA with spawn.
                # DataLoader with pin_memory=True initializes CUDA when its iterator is constructed.
                # For windows, pin_memory sometimes causes CUDA oom.
                continue

            # `exit_method` controls the way the loader process ends.
            #   - `*_kill` means that `*` is killed by OS.
            #   - `*_error` means that `*` raises an error.
            #   - `None` means that no error happens.
            # In all cases, all processes should end properly.
            if use_workers:
                exit_methods = [None, 'loader_error', 'loader_kill', 'worker_error', 'worker_kill']
            else:
                exit_methods = [None, 'loader_error', 'loader_kill']

            for exit_method in exit_methods:
                if exit_method == 'worker_kill':
                    # FIXME: This sometimes hangs. See #16608.
                    continue

                desc = []
                desc.append('is_iterable_dataset={}'.format(is_iterable_dataset))
                desc.append('use_workers={}'.format(use_workers))
                desc.append('pin_memory={}'.format(pin_memory))
                desc.append('hold_iter_reference={}'.format(hold_iter_reference))
                desc.append('exit_method={}'.format(exit_method))
                desc = 'test_proper_exit with ' + ', '.join(desc)

                # Event that the loader process uses to signal testing process
                # that various things are setup, including that the worker pids
                # are specified in `worker_pids` array.
                loader_setup_event = mp.Event()

                # Event that this process has finished setting up, and the
                # loader process can now proceed to trigger error events or
                # finish normally.
                tester_setup_event = mp.Event()

                loader_p = ErrorTrackingProcess(target=_test_proper_exit,
                                                args=(is_iterable_dataset, use_workers, pin_memory,
                                                      exit_method, hold_iter_reference,
                                                      loader_setup_event, tester_setup_event),
                                                disable_stderr=False)
                loader_p.start()
                loader_psutil_p = psutil.Process(loader_p.pid)

                # Wait for loader process to set everything up, e.g., starting
                # workers.
                loader_setup_event.wait(timeout=JOIN_TIMEOUT)
                if not loader_setup_event.is_set():
                    fail_msg = desc + ': loader process failed to setup within given time'
                    if loader_p.exception is not None:
                        fail_msg += ', and had exception {}'.format(loader_p.exception)
                    elif not loader_p.is_alive():
                        fail_msg += ', and exited with code {} but had no exception'.format(loader_p.exitcode)
                    else:
                        fail_msg += ', and is still alive.'
                    if loader_p.is_alive():
                        # this may kill the process, needs to run after the above lines
                        loader_p.print_traces_of_all_threads()
                    self.fail(fail_msg)

                # We are certain that the workers have started now.
                worker_psutil_ps = loader_psutil_p.children()

                def fail(reason):
                    report_psutil_attrs = ['pid', 'name', 'cpu_times', 'io_counters',
                                           'memory_full_info', 'num_ctx_switches',
                                           'open_files', 'threads', 'status',
                                           'nice', 'ionice']
                    if reason is None:
                        err_msg = desc
                    else:
                        err_msg = '{}: {}'.format(desc, reason)
                    err_msg += '\nLoader info:\n\t'
                    if loader_psutil_p.is_running():
                        err_msg += str(loader_psutil_p.as_dict(attrs=report_psutil_attrs))
                        # this may kill the process, needs to run after the above line
                        loader_p.print_traces_of_all_threads()
                    else:
                        err_msg += 'exited with code {}'.format(loader_p.exitcode)
                    if use_workers:
                        err_msg += '\nWorker(s) info:'
                        for idx, worker_psutil_p in enumerate(worker_psutil_ps):
                            err_msg += '\n\tWorker {}:\n\t\t'.format(idx)
                            if worker_psutil_p.is_running():
                                err_msg += str(worker_psutil_p.as_dict(attrs=report_psutil_attrs))
                                # this may kill the process, needs to run after the above line
                                print_traces_of_all_threads(worker_psutil_p.pid)
                            else:
                                err_msg += 'exited with unknown code'
                    self.fail(err_msg)

                tester_setup_event.set()

                try:
                    loader_p.join(JOIN_TIMEOUT + MP_STATUS_CHECK_INTERVAL)
                    if loader_p.is_alive():
                        fail_reason = 'loader process did not terminate'
                        if loader_p.exception is not None:
                            fail(fail_reason + ', and had exception {}'.format(loader_p.exception))
                        else:
                            fail(fail_reason + ', and had no exception')
                    _, alive = psutil.wait_procs(worker_psutil_ps, timeout=(MP_STATUS_CHECK_INTERVAL + JOIN_TIMEOUT))
                    if len(alive) > 0:
                        self.fail(get_fail_msg('worker process (pid(s) {}) did not terminate'.format(
                            ', '.join(str(p.pid) for p in alive))))
                    if exit_method is None:
                        if loader_p.exitcode != 0:
                            fail('loader process had nonzero exitcode {}'.format(loader_p.exitcode))
                    else:
                        if loader_p.exitcode == 0:
                            fail('loader process had zero exitcode')
                        if exit_method == 'loader_error':
                            if not isinstance(loader_p.exception, RuntimeError) or \
                                    'Loader error' not in str(loader_p.exception):
                                fail('loader process did not raise expected exception, but had {}'.format(
                                    loader_p.exception))
                        elif exit_method == 'worker_kill':
                            if isinstance(loader_p.exception, RuntimeError):
                                if 'DataLoader worker (pid' not in str(loader_p.exception):
                                    fail('loader process did not raise expected exception, but had {}'.format(
                                        loader_p.exception))
                            elif PY3 and isinstance(loader_p.exception, ConnectionRefusedError):
                                # Sometimes, when the worker is being killed and is freeing its
                                # resources, the unpickling in loader process will be met an
                                # a `ConnectionRefusedError` as it can not open a socket to receive
                                # resource. In such cases, the worker may not have fully exited,
                                # and the loader can't know this via `is_alive` check or `SIGCHLD`
                                # handler. So we permit this as an allowed error as well.
                                # After all, we are happy as long as it terminates.
                                pass
                            elif not PY3 and isinstance(loader_p.exception, OSError):
                                # Same reasoning as the above if-block for Py2,
                                # where ConnectionRefusedError isn't a thing.
                                if loader_p.exception.errno != errno.ECONNREFUSED:
                                    fail('loader process did not raise expected exception, but had {}'.format(
                                        loader_p.exception))
                            else:
                                fail('loader process did not raise expected exception, but had {}'.format(
                                    loader_p.exception))
                        elif exit_method == 'worker_error':
                            if not isinstance(loader_p.exception, RuntimeError) or \
                                    'Worker error' not in str(loader_p.exception):
                                fail('loader process did not raise expected exception, but had {}'.format(
                                    loader_p.exception))
                finally:
                    loader_p.terminate()

    def test_len(self):
        def check_len(dl, expected):
            self.assertEqual(len(dl), expected)
            n = 0
            for _ in dl:
                n += 1
            self.assertEqual(n, expected)
        check_len(self.dataset, 100)
        check_len(DataLoader(self.dataset, batch_size=2), 50)
        check_len(DataLoader(self.dataset, batch_size=3), 34)

    @unittest.skipIf(not TEST_NUMPY, "numpy unavailable")
    def test_numpy_scalars(self):
        import numpy as np

        class ScalarDataset(torch.utils.data.Dataset):
            def __init__(self, dtype):
                self.dtype = dtype

            def __getitem__(self, i):
                return self.dtype()

            def __len__(self):
                return 4

        dtypes = {
            np.float64: torch.DoubleTensor,
            np.float32: torch.FloatTensor,
            np.float16: torch.HalfTensor,
            np.int64: torch.LongTensor,
            np.int32: torch.IntTensor,
            np.int16: torch.ShortTensor,
            np.int8: torch.CharTensor,
            np.uint8: torch.ByteTensor,
        }
        for dt, tt in dtypes.items():
            dset = ScalarDataset(dt)
            loader = DataLoader(dset, batch_size=2)
            batch = next(iter(loader))
            self.assertIsInstance(batch, tt)

    def test_default_collate_dtype(self):
        arr = [1, 2, -1]
        collated = _utils.collate.default_collate(arr)
        self.assertEqual(collated, torch.tensor(arr))
        self.assertEqual(collated.dtype, torch.int64)

        arr = [1.1, 2.3, -0.9]
        collated = _utils.collate.default_collate(arr)
        self.assertEqual(collated, torch.tensor(arr))
        self.assertEqual(collated.dtype, torch.float64)

        arr = [True, False]
        collated = _utils.collate.default_collate(arr)
        self.assertEqual(collated, torch.tensor(arr))
        self.assertEqual(collated.dtype, torch.bool)

        # Should be a no-op
        arr = ['a', 'b', 'c']
        self.assertEqual(arr, _utils.collate.default_collate(arr))

    @unittest.skipIf(not TEST_NUMPY, "numpy unavailable")
    def test_default_collate_bad_numpy_types(self):
        import numpy as np

        # Should be a no-op
        arr = np.array(['a', 'b', 'c'])
        self.assertEqual(arr, _utils.collate.default_collate(arr))

        arr = np.array([[['a', 'b', 'c']]])
        self.assertRaises(TypeError, lambda: _utils.collate.default_collate(arr))

        arr = np.array([object(), object(), object()])
        self.assertRaises(TypeError, lambda: _utils.collate.default_collate(arr))

        arr = np.array([[[object(), object(), object()]]])
        self.assertRaises(TypeError, lambda: _utils.collate.default_collate(arr))

    @unittest.skipIf(not TEST_NUMPY, "numpy unavailable")
    def test_default_collate_shared_tensor(self):
        import numpy as np
        t_in = torch.zeros(1)
        n_in = np.zeros(1)

        self.assertEqual(t_in.is_shared(), False)

        self.assertEqual(_utils.collate.default_collate([t_in]).is_shared(), False)
        self.assertEqual(_utils.collate.default_collate([n_in]).is_shared(), False)

        # FIXME: fix the following hack that makes `default_collate` believe
        #        that it is in a worker process (since it tests
        #        `get_worker_info() != None`), even though it is not.
        old = _utils.worker._worker_info
        try:
            _utils.worker._worker_info = 'x'
            self.assertEqual(_utils.collate.default_collate([t_in]).is_shared(), True)
            self.assertEqual(_utils.collate.default_collate([n_in]).is_shared(), True)
        finally:
            _utils.worker._worker_info = old


class StringDataset(Dataset):
    def __init__(self):
        self.s = '12345'

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

    def __getitem__(self, ndx):
        return (self.s[ndx], ndx)


@unittest.skipIf(
    TEST_WITH_TSAN,
    "Fails with TSAN with the following error: starting new threads after multi-threaded "
    "fork is not supported. Dying (set die_after_fork=0 to override)")
class TestStringDataLoader(TestCase):
    def setUp(self):
        super(TestStringDataLoader, self).setUp()
        self.dataset = StringDataset()

    @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
    def test_shuffle_pin_memory(self):
        loader = DataLoader(self.dataset, batch_size=2, shuffle=True, num_workers=4, pin_memory=True)
        for (s, n) in loader:
            self.assertIsInstance(s[0], str)
            self.assertTrue(n.is_pinned())


class DictDataset(Dataset):
    def __len__(self):
        return 4

    def __getitem__(self, ndx):
        return {
            'a_tensor': torch.Tensor(4, 2).fill_(ndx),
            'another_dict': {
                'a_number': ndx,
            },
        }


@unittest.skipIf(
    TEST_WITH_TSAN,
    "Fails with TSAN with the following error: starting new threads after multi-threaded "
    "fork is not supported. Dying (set die_after_fork=0 to override)")
class TestDictDataLoader(TestCase):
    def setUp(self):
        super(TestDictDataLoader, self).setUp()
        self.dataset = DictDataset()

    def test_sequential_batch(self):
        loader = DataLoader(self.dataset, batch_size=2, shuffle=False)
        batch_size = loader.batch_size
        for i, sample in enumerate(loader):
            idx = i * batch_size
            self.assertEqual(set(sample.keys()), {'a_tensor', 'another_dict'})
            self.assertEqual(set(sample['another_dict'].keys()), {'a_number'})

            t = sample['a_tensor']
            self.assertEqual(t.size(), torch.Size([batch_size, 4, 2]))
            self.assertTrue((t[0] == idx).all())
            self.assertTrue((t[1] == idx + 1).all())

            n = sample['another_dict']['a_number']
            self.assertEqual(n.size(), torch.Size([batch_size]))
            self.assertEqual(n[0], idx)
            self.assertEqual(n[1], idx + 1)

    @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
    def test_pin_memory(self):
        loader = DataLoader(self.dataset, batch_size=2, pin_memory=True)
        for sample in loader:
            self.assertTrue(sample['a_tensor'].is_pinned())
            self.assertTrue(sample['another_dict']['a_number'].is_pinned())


class NamedTupleDataset(Dataset):
    from collections import namedtuple
    Batch = namedtuple('Batch', ['data', 'label'])
    Data = namedtuple('Data', ['positive', 'negative'])

    def __len__(self):
        return 4

    def __getitem__(self, ndx):
        return self.Batch(data=self.Data(positive=ndx, negative=-ndx),
                          label=str(ndx))


@unittest.skipIf(
    TEST_WITH_TSAN,
    "Fails with TSAN with the following error: starting new threads after multi-threaded "
    "fork is not supported. Dying (set die_after_fork=0 to override)")
class TestNamedTupleDataLoader(TestCase):
    def setUp(self):
        super(TestNamedTupleDataLoader, self).setUp()
        self.dataset = NamedTupleDataset()

    @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
    def test_collate_and_pin_memory_with_namedtuple(self):
        loader = DataLoader(self.dataset, batch_size=2, pin_memory=True)
        for batch in loader:
            self.assertIsInstance(batch, NamedTupleDataset.Batch)
            self.assertIsInstance(batch.data, NamedTupleDataset.Data)


class SimpleCustomBatch(object):
    def __init__(self, data):
        transposed_data = list(zip(*data))
        self.inp = torch.stack(transposed_data[0], 0)
        self.tgt = torch.stack(transposed_data[1], 0)

    def pin_memory(self):
        self.inp = self.inp.pin_memory()
        self.tgt = self.tgt.pin_memory()
        return self

    def is_pinned(self):
        return self.inp.is_pinned() and self.tgt.is_pinned()


def collate_wrapper(batch):
    return SimpleCustomBatch(batch)


def collate_into_packed_sequence(batch):
    data = torch.stack([sample[0] for sample in batch], 1)
    t, b = data.size()
    lengths = torch.randint(1, t, size=(b,), dtype=torch.int64)
    return torch.nn.utils.rnn.pack_padded_sequence(data, lengths, enforce_sorted=False)


def collate_into_packed_sequence_batch_first(batch):
    data = torch.stack([sample[0] for sample in batch], 0)
    b, t = data.size()
    lengths = torch.randint(1, t, size=(b,), dtype=torch.int64)
    return torch.nn.utils.rnn.pack_padded_sequence(data, lengths, batch_first=True, enforce_sorted=False)


@unittest.skipIf(
    TEST_WITH_TSAN,
    "Fails with TSAN with the following error: starting new threads after multi-threaded "
    "fork is not supported. Dying (set die_after_fork=0 to override)")
class TestCustomPinFn(TestCase):
    def setUp(self):
        super(TestCustomPinFn, self).setUp()
        inps = torch.arange(10 * 5, dtype=torch.float32).view(10, 5)
        tgts = torch.arange(10 * 5, dtype=torch.float32).view(10, 5)
        self.dataset = TensorDataset(inps, tgts)

    @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
    @skipIfRocm
    def test_custom_batch_pin(self):
        test_cases = [
            (collate_wrapper, SimpleCustomBatch),
            (collate_into_packed_sequence, torch.nn.utils.rnn.PackedSequence),
            (collate_into_packed_sequence_batch_first, torch.nn.utils.rnn.PackedSequence),
        ]
        for collate_fn, elem_cls in test_cases:
            loader = DataLoader(self.dataset, batch_size=2, collate_fn=collate_fn,
                                pin_memory=True)
            for sample in loader:
                self.assertIsInstance(sample, elem_cls)
                self.assertTrue(sample.is_pinned())

    @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
    @skipIfRocm
    def test_custom_batch_pin_worker(self):
        test_cases = [
            (collate_wrapper, SimpleCustomBatch),
            (collate_into_packed_sequence, torch.nn.utils.rnn.PackedSequence),
            (collate_into_packed_sequence_batch_first, torch.nn.utils.rnn.PackedSequence),
        ]
        for collate_fn, elem_cls in test_cases:
            loader = DataLoader(self.dataset, batch_size=2, collate_fn=collate_fn,
                                pin_memory=True, num_workers=1)
            for sample in loader:
                self.assertIsInstance(sample, elem_cls)
                self.assertTrue(sample.is_pinned())


class TestWorkerQueueDataset(Dataset):
    def __init__(self, data):
        self.data = data
        self.worker_id = None

    def worker_init_fn(self, worker_id):
        self.worker_id = worker_id

    def __getitem__(self, item):
        return self.worker_id, self.data[item]

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


@unittest.skipIf(
    TEST_WITH_TSAN,
    "Fails with TSAN with the following error: starting new threads after multi-threaded "
    "fork is not supported. Dying (set die_after_fork=0 to override)")
class TestIndividualWorkerQueue(TestCase):
    def setUp(self):
        super(TestIndividualWorkerQueue, self).setUp()
        self.dataset = TestWorkerQueueDataset([i for i in range(128)])

    def _run_ind_worker_queue_test(self, batch_size, num_workers):
        loader = DataLoader(
            self.dataset, batch_size=batch_size, shuffle=False, num_workers=num_workers,
            worker_init_fn=self.dataset.worker_init_fn
        )
        current_worker_idx = 0
        for i, (worker_ids, sample) in enumerate(loader):
            self.assertEqual(worker_ids.tolist(), [current_worker_idx] * batch_size)
            self.assertEqual(sample.tolist(), [j for j in range(i * batch_size, (i + 1) * batch_size)])
            current_worker_idx += 1
            if current_worker_idx == num_workers:
                current_worker_idx = 0

    def test_ind_worker_queue(self):
        for batch_size in (8, 16, 32, 64):
            for num_workers in range(1, 6):
                self._run_ind_worker_queue_test(batch_size=batch_size, num_workers=num_workers)


if __name__ == '__main__':
    run_tests()