README.md

# bert4sl_punct_zh_public

## 时间(time)

2024.6



## 训练数据构成(dataset)

使用高质量语料过滤而成, 收集高质量语料, 并使用PPL过滤等;

 - [chinese-poetry/chinese-poetry](https://github.com/chinese-poetry/chinese-poetry)

 - [chinese-poetry/huajianji](https://github.com/chinese-poetry/huajianji)

 - [garychowcmu/daizhigev20](https://github.com/garychowcmu/daizhigev20)

 - [yangjianxin1/Firefly](https://github.com/yangjianxin1/Firefly)

 - [学习强国428万数据](https://huggingface.co/datasets/Macropodus/xuexiqiangguo_428w); 国内源[Macropodus/xuexiqiangguo_428w](https://hf-mirror.com/datasets/Macropodus/xuexiqiangguo_428w)

 - [xi_talk40万](https://huggingface.co/datasets/Papersnake/xi_talk); 国内源[Papersnake/xi_talk](https://hf-mirror.com/datasets/Papersnake/xi_talk)

 - [qwen-7b生成的100万好句]

 - [人民日报语料2000万]



## 训练说明

每种标点的最大句子数为10万, 总计500万训练句子, 训练3epoch;