如何直接从列表或字典创建 torchtext.data.TabularDataset

Question

torchtext.data.TabularDataset 可以从 TSV/JSON/CSV 文件创建，然后可用于从 Glove、FastText 或任何其他嵌入构建词汇表。但我的要求是直接从 list 或 dict 创建 torchtext.data.TabularDataset。

当前通过读取 TSV 文件实现的代码

self.RAW = data.RawField()
self.TEXT = data.Field(batch_first=True)
self.LABEL = data.Field(sequential=False, unk_token=None)


self.train, self.dev, self.test = data.TabularDataset.splits(
    path='.data/quora',
    train='train.tsv',
    validation='dev.tsv',
    test='test.tsv',
    format='tsv',
    fields=[('label', self.LABEL),
            ('q1', self.TEXT),
            ('q2', self.TEXT),
            ('id', self.RAW)])


self.TEXT.build_vocab(self.train, self.dev, self.test, vectors=GloVe(name='840B', dim=300))
self.LABEL.build_vocab(self.train)


sort_key = lambda x: data.interleave_keys(len(x.q1), len(x.q2))


self.train_iter, self.dev_iter, self.test_iter = \
    data.BucketIterator.splits((self.train, self.dev, self.test),
                               batch_sizes=[args.batch_size] * 3,
                               device=args.gpu,
                               sort_key=sort_key)

这是从文件中读取数据的当前工作代码。因此，为了直接从 List/Dict 创建数据集，我尝试了 Examples.fromDict 或 Examples.fromList 等内置函数，但在进入最后一个 for 循环时，它抛出了一个错误 AttributeError: 'BucketIterator' object has no attribute 'q1'

Answer 1

它要求我编写一个自己的 class 继承数据集 class 并在 torchtext.data.TabularDataset class.

中进行少量修改

class TabularDataset_From_List(data.Dataset):

    def __init__(self, input_list, format, fields, skip_header=False, **kwargs):
        make_example = {
            'json': Example.fromJSON, 'dict': Example.fromdict,
            'tsv': Example.fromTSV, 'csv': Example.fromCSV}[format.lower()]

        examples = [make_example(item, fields) for item in input_list]

        if make_example in (Example.fromdict, Example.fromJSON):
            fields, field_dict = [], fields
            for field in field_dict.values():
                if isinstance(field, list):
                    fields.extend(field)
                else:
                    fields.append(field)

        super(TabularDataset_From_List, self).__init__(examples, fields, **kwargs)

    @classmethod
    def splits(cls, path=None, root='.data', train=None, validation=None,
               test=None, **kwargs):
        if path is None:
            path = cls.download(root)
        train_data = None if train is None else cls(
            train, **kwargs)
        val_data = None if validation is None else cls(
            validation, **kwargs)
        test_data = None if test is None else cls(
            test, **kwargs)
        return tuple(d for d in (train_data, val_data, test_data)
                     if d is not None)

如何直接从列表或字典创建 torchtext.data.TabularDataset

How to create a torchtext.data.TabularDataset directly from a list or dict

python

dataset

pytorch

torchtext