数据集按照类的方式写作，采用自定义数据采集器，出现数据无法加载现象

[niushao12]

bug描述 Describe the Bug

按照paddle提供代码进行数据集前处理图类构建，通过自定义数据采集器，训练测试数据无法load。

paddlepaddle ==2.6.1

以下是整个代码文本

import paddle
import numpy as np
from random import Random
from tqdm import trange, tqdm

class MoleculeDatapoint:
def init(self, smiles, targets, features):
self.smiles = smiles
self.targets = targets
self.features = features

class MoleculeDataset(paddle.io.Dataset):
def init(self, data):
self._data = data
self._batch_graph = None
self._random = Random()

def __len__(self) ->int:
    return len(self._data)

def __getitem__(self, item):
    print(item, self._data[item].smiles)
    return self._data[item]

class MoleculeSampler(paddle.io.Sampler):
def init(self, dataset, class_balance: bool=False,
shuffle: bool=False, seed: int=0):
super(paddle.io.Sampler, self).init()
self.dataset = dataset
self.class_balance = class_balance
self.shuffle = shuffle
self._random = Random(seed)
if self.class_balance:
indices = np.arange(len(dataset))
has_active = np.array([any(target == 1 for target in datapoint.
targets) for datapoint in dataset])
self.positive_indices = indices[has_active].tolist()
self.negative_indices = indices[~has_active].tolist()
self.length = 2 * min(len(self.positive_indices), len(self.
negative_indices))
else:
self.positive_indices = self.negative_indices = None
self.length = len(self.dataset)

def __iter__(self):
    if self.class_balance:
        if self.shuffle:
            self._random.shuffle(self.positive_indices)
            self._random.shuffle(self.negative_indices)
        indices = [index for pair in zip(self.positive_indices, self.
            negative_indices) for index in pair]
    else:
        indices = list(range(len(self.dataset)))
        if self.shuffle:
            self._random.shuffle(indices)
    return iter(indices)

def __len__(self):
    return self.length

class DataLoader(paddle.io.DataLoader):
def init(self,
dataset,
batch_size=1,
shuffle=False,
sampler=None,
batch_sampler=None,
num_workers=0,
collate_fn=None,
pin_memory=False,
drop_last=False,
timeout=0,
worker_init_fn=None,
multiprocessing_context=None,
generator=None):

    if isinstance(dataset[0], (tuple, list)):
        return_list = True
    else:
        return_list = False

    super().__init__(
        dataset,
        feed_list=None,
        places=None,
        return_list=return_list,
        batch_sampler=batch_sampler,
        batch_size=batch_size,
        shuffle=shuffle,
        drop_last=drop_last,
        collate_fn=collate_fn,
        num_workers=num_workers,
        use_buffer_reader=True,
        use_shared_memory=False,
        timeout=timeout,
        worker_init_fn=worker_init_fn)
    if sampler is not None:
        self.batch_sampler.sampler = sampler

def construct_molecule_batch(data):
data = MoleculeDataset(data)
return data

class MoleculeDataLoader(DataLoader):
def init(self, dataset, batch_size: int=50,
num_workers: int=8, class_balance: bool=False, shuffle: bool=False,
seed: int=0):
self._dataset = dataset
self._batch_size = batch_size
self._num_workers = num_workers
self._class_balance = class_balance
self._shuffle = shuffle
self._seed = seed
self._context = None
self._timeout = 0
self._sampler = MoleculeSampler(dataset=self._dataset,
class_balance=self._class_balance, shuffle=self._shuffle, seed=
self._seed)
super(MoleculeDataLoader, self).init(dataset=self._dataset,
batch_size=self._batch_size, sampler=self._sampler, num_workers
=self._num_workers, collate_fn=construct_molecule_batch,
multiprocessing_context=self._context, timeout=self._timeout)

def __iter__(self):
    return super(MoleculeDataLoader, self).__iter__()

def chemprop_build_data_loader(
smiles: list[str],
fingerprints: np.ndarray = None,
properties: list[int] = None ,
shuffle: bool = False,
num_workers: int = 0
):

if fingerprints is None:
    fingerprints = [None] * len(smiles)

if properties is None:
    properties = [None] * len(smiles)
else:
    properties = [[float(prop)] for prop in properties]

return MoleculeDataLoader(
    dataset=MoleculeDataset([
        MoleculeDatapoint(
            smiles=[smiles],
            targets=prop,
            features=fingerprint,
        ) for smiles, fingerprint, prop in zip(smiles, fingerprints, properties)
    ]),
    batch_size = 4,
    num_workers=num_workers,
    shuffle=True
)

def main():
train_smiles=[384.27,405.311,386.286,356.26,384.27,463.441,236.109,352.316,379.367,275.106,507.472,292.14,98.145,312.453,284.44,570.896,326.477,654.97,223.747,308.636,195.693,350.48,310.437,298.426,298.426,312.453,284.443,337.463,296.41,310.437,364.529,310.481,369.505,340.463,384.516,368.524,466.456,500.629,282.512,204.357]
train_fingerprints=[2.6355,4.39088,1.4796,2.4663,1.496,4.3989,0.9992,3.9613,4.44214,-0.4514,4.5354,1.7634,0.7806,3.8826,3.0149,6.5826,3.5857,7.7242,2.6044,3.7491,2.1734,3.9192,3.6586,3.6366,3.4925,3.8826,4.3135,4.221,3.6126,3.9156,5.2284,4.8697,4.7145,4.0633,4.4259,5.15,2.63012,2.5925,5.974,5.0354]
train_properties=[384.27,405.311,386.286,356.26,384.27,463.441,236.109,352.316,379.367,275.106,507.472,292.14,98.145,312.453,284.44,570.896,326.477,654.97,223.747,308.636,195.693,350.48,310.437,298.426,298.426,312.453,284.443,337.463,296.41,310.437,364.529,310.481,369.505,340.463,384.516,368.524,466.456,500.629,282.512,204.357]

# Build data loaders
train_data_loader = chemprop_build_data_loader(
    smiles=train_smiles,
    fingerprints=train_fingerprints,
    properties=train_properties,
    shuffle=True,
    num_workers=0
)

print(len(train_data_loader))

for batch in tqdm(train_data_loader, total=len(train_data_loader), leave=False):#：这里出现了bug，无法进行的数据，数据长度没有问题，在torch下完全可以加载数据。
    batch: MoleculeDataset
    smile = batch.smiles()
    print("batch", smile)

if name == 'main':
main()

其他补充信息 Additional Supplementary Information

No response

[liaoxin2]

您好，我初步排查了一下，该问题可能是由于construct_molecule_batch函数中的data不是张量造成的，可以参看https://github.com/PaddlePaddle/Paddle/issues/41883，该问题我已反馈，后续版本应该会得到解决