rapidnlp-datasets

Data pipelines for both TensorFlow and PyTorch !

If you want to load public datasets, try:

• tensorflow/datasets
• huggingface/datasets

If you want to load local, personal dataset with minimized boilerplate, use rapidnlp-datasets!

installation

pip install -U rapidnlp-datasets

If you work with PyTorch, you should install PyTorch first.

If you work with TensorFlow, you should install TensorFlow first.

Usage

Here are few examples to show you how to use this library.

• QuickStart: Sequence Classification Task
• QuickStart: Question Answering Task
• QuickStart: Token Classification Task
• QuickStart: Masked Language Model Task
• QuickStart: SimCSE(Sentence Embedding)

sequence-classification-quickstart

In PyTorch,

>>> import torch
>>> from rapidnlp_datasets.pt import DatasetForSequenceClassification
>>> dataset = DatasetForSequenceClassification.from_jsonl_files(
        input_files=["testdata/sequence_classification.jsonl"],
        vocab_file="testdata/vocab.txt",
    )
>>> dataloader = torch.utils.data.DataLoader(dataset, shuffle=True, batch_size=32, collate_fn=dataset.batch_padding_collate)
>>> for idx, batch in enumerate(dataloader):
...     print("No.{} batch: \n{}".format(idx, batch))
... 

In TensorFlow,

>>> from rapidnlp_datasets.tf import TFDatasetForSequenceClassifiation
>>> dataset, d = TFDatasetForSequenceClassifiation.from_jsonl_files(
        input_files=["testdata/sequence_classification.jsonl"],
        vocab_file="testdata/vocab.txt",
        return_self=True,
    )
>>> for idx, batch in enumerate(iter(dataset)):
...     print("No.{} batch: \n{}".format(idx, batch))
... 

Especially, you can save dataset to tfrecord format when working with TensorFlow, and then build dataset from tfrecord files directly!

>>> d.save_tfrecord("testdata/sequence_classification.tfrecord")
2021-12-08 14:52:41,295    INFO             utils.py  128] Finished to write 2 examples to tfrecords.
>>> dataset = TFDatasetForSequenceClassifiation.from_tfrecord_files("testdata/sequence_classification.tfrecord")
>>> for idx, batch in enumerate(iter(dataset)):
...     print("No.{} batch: \n{}".format(idx, batch))
... 

question-answering-quickstart

In PyTorch:

>>> import torch
>>> from rapidnlp_datasets.pt import DatasetForQuestionAnswering
>>>
>>> dataset = DatasetForQuestionAnswering.from_jsonl_files(
        input_files="testdata/qa.jsonl",
        vocab_file="testdata/vocab.txt",
    )
>>> dataloader = torch.utils.data.DataLoader(dataset, shuffle=True, batch_size=32, collate_fn=dataset.batch_padding_collate)
>>> for idx, batch in enumerate(dataloader):
...     print("No.{} batch: \n{}".format(idx, batch))
... 

In TensorFlow,

>>> from rapidnlp_datasets.tf import TFDatasetForQuestionAnswering
>>> dataset, d = TFDatasetForQuestionAnswering.from_jsonl_files(
        input_files="testdata/qa.jsonl",
        vocab_file="testdata/vocab.txt",
        return_self=True,
    )
2021-12-08 15:09:06,747    INFO question_answering_dataset.py  101] Read 3 examples in total.
>>> for idx, batch in enumerate(iter(dataset)):
        print()
        print("NO.{} batch: \n{}".format(idx, batch))
... 

Especially, you can save dataset to tfrecord format when working with TensorFlow, and then build dataset from tfrecord files directly!

>>> d.save_tfrecord("testdata/qa.tfrecord")
2021-12-08 15:09:31,329    INFO             utils.py  128] Finished to write 3 examples to tfrecords.
>>> dataset = TFDatasetForQuestionAnswering.from_tfrecord_files(
        "testdata/qa.tfrecord",
        batch_size=32,
        padding="batch",
    )
>>> for idx, batch in enumerate(iter(dataset)):
        print()
        print("NO.{} batch: \n{}".format(idx, batch))
... 

token-classification-quickstart

masked-language-models-quickstart

simcse-quickstart