Python mmcv.parallel.collate() Examples

def inference_detector(model, img):
    """Inference image(s) with the detector.

    Args:
        model (nn.Module): The loaded detector.
        imgs (str/ndarray or list[str/ndarray]): Either image files or loaded
            images.

    Returns:
        If imgs is a str, a generator will be returned, otherwise return the
        detection results directly.
    """
    cfg = model.cfg
    device = next(model.parameters()).device  # model device
    # build the data pipeline
    test_pipeline = [LoadImage()] + cfg.data.test.pipeline[1:]
    test_pipeline = Compose(test_pipeline)
    # prepare data
    data = dict(img=img)
    data = test_pipeline(data)
    data = scatter(collate([data], samples_per_gpu=1), [device])[0]
    # forward the model
    with torch.no_grad():
        result = model(return_loss=False, rescale=True, **data)
    return result 

def _data_func(data, device_id):
    data = scatter(collate([data], samples_per_gpu=1), [device_id])[0]
    return dict(return_loss=False, rescale=True, **data) 

def inference_detector(model, img):
    """Inference image(s) with the detector.

    Args:
        model (nn.Module): The loaded detector.
        imgs (str/ndarray or list[str/ndarray]): Either image files or loaded
            images.

    Returns:
        If imgs is a str, a generator will be returned, otherwise return the
        detection results directly.
    """
    cfg = model.cfg
    device = next(model.parameters()).device  # model device
    # build the data pipeline
    test_pipeline = [LoadImage()] + cfg.data.test.pipeline[1:]
    test_pipeline = Compose(test_pipeline)
    # prepare data
    data = dict(img=img)
    data = test_pipeline(data)
    data = collate([data], samples_per_gpu=1)
    if next(model.parameters()).is_cuda:
        # scatter to specified GPU
        data = scatter(data, [device])[0]
    else:
        # Use torchvision ops for CPU mode instead
        for m in model.modules():
            if isinstance(m, (RoIPool, RoIAlign)):
                if not m.aligned:
                    # aligned=False is not implemented on CPU
                    # set use_torchvision on-the-fly
                    m.use_torchvision = True
        warnings.warn('We set use_torchvision=True in CPU mode.')
        # just get the actual data from DataContainer
        data['img_metas'] = data['img_metas'][0].data

    # forward the model
    with torch.no_grad():
        result = model(return_loss=False, rescale=True, **data)
    return result 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    shuffle = kwargs.get('shuffle', True)
    if dist:
        rank, world_size = get_dist_info()
        if shuffle:
            sampler = DistributedGroupSampler(dataset, imgs_per_gpu,
                                              world_size, rank)
        else:
            sampler = DistributedSampler(
                dataset, world_size, rank, shuffle=False)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        sampler = GroupSampler(dataset, imgs_per_gpu) if shuffle else None
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def _data_func(data, device_id):
    data = scatter(collate([data], samples_per_gpu=1), [device_id])[0]
    return dict(return_loss=False, rescale=True, **data) 

def inference_detector(model, img):
    """Inference image(s) with the detector.

    Args:
        model (nn.Module): The loaded detector.
        imgs (str/ndarray or list[str/ndarray]): Either image files or loaded
            images.

    Returns:
        If imgs is a str, a generator will be returned, otherwise return the
        detection results directly.
    """
    cfg = model.cfg
    device = next(model.parameters()).device  # model device
    # build the data pipeline
    test_pipeline = [LoadImage()] + cfg.data.test.pipeline[1:]
    test_pipeline = Compose(test_pipeline)
    # prepare data
    data = dict(img=img)
    data = test_pipeline(data)
    data = scatter(collate([data], samples_per_gpu=1), [device])[0]
    # forward the model
    with torch.no_grad():
        result = model(return_loss=False, rescale=True, **data)

    return result


# TODO: merge this method with the one in BaseDetector 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    shuffle = kwargs.get('shuffle', True)
    if dist:
        rank, world_size = get_dist_info()
        if shuffle:
            sampler = DistributedGroupSampler(dataset, imgs_per_gpu,
                                              world_size, rank)
        else:
            sampler = DistributedSampler(
                dataset, world_size, rank, shuffle=False)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        sampler = GroupSampler(dataset, imgs_per_gpu) if shuffle else None
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def inference_detector(model, img):
    """Inference image(s) with the detector.

    Args:
        model (nn.Module): The loaded detector.
        imgs (str/ndarray or list[str/ndarray]): Either image files or loaded
            images.

    Returns:
        If imgs is a str, a generator will be returned, otherwise return the
        detection results directly.
    """
    cfg = model.cfg
    device = next(model.parameters()).device  # model device
    # build the data pipeline
    test_pipeline = [LoadImage()] + cfg.data.test.pipeline[1:]
    test_pipeline = Compose(test_pipeline)
    # prepare data
    data = dict(img=img)
    data = test_pipeline(data)
    data = scatter(collate([data], samples_per_gpu=1), [device])[0]
    # forward the model
    with torch.no_grad():
        result = model(return_loss=False, rescale=True, **data)

    return result


# TODO: merge this method with the one in BaseDetector 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    shuffle = kwargs.get('shuffle', True)
    if dist:
        rank, world_size = get_dist_info()
        if shuffle:
            sampler = DistributedGroupSampler(dataset, imgs_per_gpu,
                                              world_size, rank)
        else:
            sampler = DistributedSampler(
                dataset, world_size, rank, shuffle=False)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        sampler = GroupSampler(dataset, imgs_per_gpu) if shuffle else None
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    shuffle = kwargs.get('shuffle', True)
    if dist:
        rank, world_size = get_dist_info()
        if shuffle:
            sampler = DistributedGroupSampler(dataset, imgs_per_gpu,
                                              world_size, rank)
        else:
            sampler = DistributedSampler(dataset,
                                         world_size,
                                         rank,
                                         shuffle=False)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        sampler = GroupSampler(dataset, imgs_per_gpu) if shuffle else None
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(dataset,
                             batch_size=batch_size,
                             sampler=sampler,
                             num_workers=num_workers,
                             collate_fn=partial(collate,
                                                samples_per_gpu=imgs_per_gpu),
                             pin_memory=False,
                             **kwargs)

    return data_loader 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    shuffle = kwargs.get('shuffle', True)
    if dist:
        rank, world_size = get_dist_info()
        if shuffle:
            sampler = DistributedGroupSampler(dataset, imgs_per_gpu, world_size, rank)
        else:
            sampler = DistributedSampler(dataset, world_size, rank, shuffle=False)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        if not kwargs.get('shuffle', True):
            sampler = None
        else:
            sampler = GroupSampler(dataset, imgs_per_gpu)
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def _data_func(data, device_id):
    data = scatter(collate([data], samples_per_gpu=1), [device_id])[0]
    return dict(return_loss=False, rescale=True, **data) 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    if dist:
        rank, world_size = get_dist_info()
        sampler = DistributedGroupSampler(dataset, imgs_per_gpu, world_size,
                                          rank)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        if not kwargs.get('shuffle', True):
            sampler = None
        else:
            sampler = GroupSampler(dataset, imgs_per_gpu)
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def _data_func(data, device_id):
    data = scatter(collate([data], samples_per_gpu=1), [device_id])[0]
    return dict(return_loss=False, rescale=True, **data) 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    shuffle = kwargs.get('shuffle', True)
    if dist:
        rank, world_size = get_dist_info()
        if shuffle:
            sampler = DistributedGroupSampler(dataset, imgs_per_gpu,
                                              world_size, rank)
        else:
            sampler = DistributedSampler(
                dataset, world_size, rank, shuffle=False)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        sampler = GroupSampler(dataset, imgs_per_gpu) if shuffle else None
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def inference_detector(model, img):
    """Inference image(s) with the detector.

    Args:
        model (nn.Module): The loaded detector.
        imgs (str/ndarray or list[str/ndarray]): Either image files or loaded
            images.

    Returns:
        If imgs is a str, a generator will be returned, otherwise return the
        detection results directly.
    """
    cfg = model.cfg
    device = next(model.parameters()).device  # model device
    # build the data pipeline
    test_pipeline = [LoadImage()] + cfg.data.test.pipeline[1:]
    test_pipeline = Compose(test_pipeline)
    # prepare data
    data = dict(img=img)
    data = test_pipeline(data)
    data = scatter(collate([data], samples_per_gpu=1), [device])[0]
    # forward the model
    with torch.no_grad():
        result = model(return_loss=False, rescale=True, **data)

    return result


# TODO: merge this method with the one in BaseDetector 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    shuffle = kwargs.get('shuffle', True)
    if dist:
        rank, world_size = get_dist_info()
        if shuffle:
            sampler = DistributedGroupSampler(dataset, imgs_per_gpu,
                                              world_size, rank)
        else:
            sampler = DistributedSampler(
                dataset, world_size, rank, shuffle=False)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        sampler = GroupSampler(dataset, imgs_per_gpu) if shuffle else None
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    if dist:
        rank, world_size = get_dist_info()
        sampler = DistributedGroupSampler(dataset, imgs_per_gpu, world_size,
                                          rank)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        if not kwargs.get('shuffle', True):
            sampler = None
        else:
            sampler = GroupSampler(dataset, imgs_per_gpu)
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def _data_func(data, device_id):
    data = scatter(collate([data], samples_per_gpu=1), [device_id])[0]
    return dict(return_loss=False, rescale=True, **data) 

def _data_func(data, device_id):
    data = scatter(collate([data], samples_per_gpu=1), [device_id])[0]
    return dict(return_loss=False, rescale=True, **data) 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    shuffle = kwargs.get('shuffle', True)
    if dist:
        rank, world_size = get_dist_info()
        if shuffle:
            sampler = DistributedGroupSampler(dataset, imgs_per_gpu,
                                              world_size, rank)
        else:
            sampler = DistributedSampler(
                dataset, world_size, rank, shuffle=False)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        sampler = GroupSampler(dataset, imgs_per_gpu) if shuffle else None
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def async_inference_detector(model, img):
    """Async inference image(s) with the detector.

    Args:
        model (nn.Module): The loaded detector.
        imgs (str/ndarray or list[str/ndarray]): Either image files or loaded
            images.

    Returns:
        Awaitable detection results.
    """
    cfg = model.cfg
    device = next(model.parameters()).device  # model device
    # build the data pipeline
    test_pipeline = [LoadImage()] + cfg.data.test.pipeline[1:]
    test_pipeline = Compose(test_pipeline)
    # prepare data
    data = dict(img=img)
    data = test_pipeline(data)
    data = scatter(collate([data], samples_per_gpu=1), [device])[0]

    # We don't restore `torch.is_grad_enabled()` value during concurrent
    # inference since execution can overlap
    torch.set_grad_enabled(False)
    result = await model.aforward_test(rescale=True, **data)
    return result 

def model_aug_test_template(cfg_file):
    # get config
    cfg = mmcv.Config.fromfile(cfg_file)
    # init model
    cfg.model.pretrained = None
    model = build_detector(cfg.model, train_cfg=None, test_cfg=cfg.test_cfg)

    # init test pipeline and set aug test
    load_cfg, multi_scale_cfg = cfg.test_pipeline
    multi_scale_cfg['flip'] = True
    multi_scale_cfg['img_scale'] = [(1333, 800), (800, 600), (640, 480)]

    load = build_from_cfg(load_cfg, PIPELINES)
    transform = build_from_cfg(multi_scale_cfg, PIPELINES)

    results = dict(
        img_prefix=osp.join(osp.dirname(__file__), '../data'),
        img_info=dict(filename='color.jpg'))
    results = transform(load(results))
    assert len(results['img']) == 6
    assert len(results['img_metas']) == 6

    results['img'] = [collate([x]) for x in results['img']]
    results['img_metas'] = [collate([x]).data[0] for x in results['img_metas']]
    # aug test the model
    model.eval()
    with torch.no_grad():
        aug_result = model(return_loss=False, rescale=True, **results)
    return aug_result 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    shuffle = kwargs.get('shuffle', True)
    if dist:
        rank, world_size = get_dist_info()
        if shuffle:
            sampler = DistributedGroupSampler(dataset, imgs_per_gpu,
                                              world_size, rank)
        else:
            sampler = DistributedSampler(
                dataset, world_size, rank, shuffle=False)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        sampler = GroupSampler(dataset, imgs_per_gpu) if shuffle else None
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def build_data_loader(
        dataset,
        imgs_per_gpu,
        workers_per_gpu,
        num_gpus=1,
        dist=True,
        **kwargs
):
    shuffle = kwargs.get('shuffle', True)
    if dist:
        rank, world_size = get_dist_info()
        if shuffle:
            sampler = DistributedGroupSampler(
                dataset, imgs_per_gpu, world_size, rank
            )
        else:
            sampler = DistributedSampler(
                dataset, world_size, rank, shuffle=False
            )
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        sampler = GroupSampler(dataset, imgs_per_gpu) if shuffle else None
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    shuffle = kwargs.get('shuffle', True)
    if dist:
        rank, world_size = get_dist_info()
        if shuffle:
            sampler = DistributedGroupSampler(dataset, imgs_per_gpu,
                                              world_size, rank)
        else:
            sampler = DistributedSampler(
                dataset, world_size, rank, shuffle=False)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        sampler = GroupSampler(dataset, imgs_per_gpu) if shuffle else None
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    shuffle = kwargs.get('shuffle', True)
    if dist:
        rank, world_size = get_dist_info()
        if shuffle:
            sampler = DistributedGroupSampler(dataset, imgs_per_gpu,
                                              world_size, rank)
        else:
            sampler = DistributedSampler(
                dataset, world_size, rank, shuffle=False)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        # 非分布式训练
        sampler = GroupSampler(dataset, imgs_per_gpu) if shuffle else None  # batch中样本的采样方式
        batch_size = num_gpus * imgs_per_gpu        # 在这里定义batch size
        num_workers = num_gpus * workers_per_gpu    # 多线程读取可以加快数据的读取速度

    # 采用pytorch内置的DataLoader方法
    # DataLoader是一个 迭代器
    # collate_fn：在数据处理中，有时会出现某个样本无法读取等问题，比如某张图片损坏。
    #             这时在_ getitem _函数中将出现异常，此时最好的解决方案即是将出错的样本剔除。
    #             如果实在是遇到这种情况无法处理，则可以返回None对象，然后在Dataloader中实现自定义的collate_fn，将空对象过滤掉。
    #             但要注意，在这种情况下dataloader返回的batch数目会少于batch_size。

    # sampler：自定义从数据集中取样本的策略，如果指定这个参数，那么shuffle必须为False
    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def async_inference_detector(model, img):
    """Async inference image(s) with the detector.

    Args:
        model (nn.Module): The loaded detector.
        imgs (str/ndarray or list[str/ndarray]): Either image files or loaded
            images.

    Returns:
        Awaitable detection results.
    """
    cfg = model.cfg
    device = next(model.parameters()).device  # model device
    # build the data pipeline
    test_pipeline = [LoadImage()] + cfg.data.test.pipeline[1:]
    test_pipeline = Compose(test_pipeline)
    # prepare data
    data = dict(img=img)
    data = test_pipeline(data)
    data = scatter(collate([data], samples_per_gpu=1), [device])[0]

    # We don't restore `torch.is_grad_enabled()` value during concurrent
    # inference since execution can overlap
    torch.set_grad_enabled(False)
    result = await model.aforward_test(rescale=True, **data)
    return result


# TODO: merge this method with the one in BaseDetector 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    if dist:
        rank, world_size = get_dist_info()
        sampler = DistributedGroupSampler(dataset, imgs_per_gpu, world_size,
                                          rank)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        if not kwargs.get('shuffle', True):
            sampler = None
        else:
            sampler = GroupSampler(dataset, imgs_per_gpu)
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader 

def build_dataloader(dataset,
                     imgs_per_gpu,
                     workers_per_gpu,
                     num_gpus=1,
                     dist=True,
                     **kwargs):
    shuffle = kwargs.get('shuffle', True)
    if dist:
        rank, world_size = get_dist_info()
        if shuffle:
            sampler = DistributedGroupSampler(dataset, imgs_per_gpu,
                                              world_size, rank)
        else:
            sampler = DistributedSampler(
                dataset, world_size, rank, shuffle=False)
        batch_size = imgs_per_gpu
        num_workers = workers_per_gpu
    else:
        sampler = GroupSampler(dataset, imgs_per_gpu) if shuffle else None
        batch_size = num_gpus * imgs_per_gpu
        num_workers = num_gpus * workers_per_gpu

    data_loader = DataLoader(
        dataset,
        batch_size=batch_size,
        sampler=sampler,
        num_workers=num_workers,
        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
        pin_memory=False,
        **kwargs)

    return data_loader