Chinese Bigbird Mini 1024

🚀 Chinese BigBird Mini 1024

This project provides a customized tokenizer based on Jieba and a pre - trained BigBird model for Chinese text processing.

🚀 Quick Start

The following is an example of using the pre - trained Chinese BigBird model and the customized Jieba tokenizer:

Basic Usage

import jieba_fast
from transformers import BertTokenizer
from transformers import BigBirdModel

class JiebaTokenizer(BertTokenizer):
    def __init__(
        self, pre_tokenizer=lambda x: jieba_fast.cut(x, HMM=False), *args, **kwargs
    ):
        super().__init__(*args, **kwargs)
        self.pre_tokenizer = pre_tokenizer

    def _tokenize(self, text, *arg, **kwargs):
        split_tokens = []
        for text in self.pre_tokenizer(text):
            if text in self.vocab:
                split_tokens.append(text)
            else:
                split_tokens.extend(super()._tokenize(text))
        return split_tokens

model = BigBirdModel.from_pretrained('Lowin/chinese-bigbird-mini-1024')
tokenizer = JiebaTokenizer.from_pretrained('Lowin/chinese-bigbird-mini-1024')

You can access the project on GitHub: https://github.com/LowinLi/chinese-bigbird

📄 License

This project is licensed under the Apache 2.0 license.