Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/ZhuiyiTechnology/roformer-v2
RoFormer升级版
https://github.com/ZhuiyiTechnology/roformer-v2
Last synced: 20 days ago
JSON representation
RoFormer升级版
- Host: GitHub
- URL: https://github.com/ZhuiyiTechnology/roformer-v2
- Owner: ZhuiyiTechnology
- License: apache-2.0
- Created: 2022-03-17T02:26:13.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2022-08-11T02:30:07.000Z (over 2 years ago)
- Last Synced: 2024-08-01T03:19:07.503Z (4 months ago)
- Language: Python
- Size: 42 KB
- Stars: 151
- Watchers: 7
- Forks: 15
- Open Issues: 5
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-chinese-ner - roformer-v2
README
# RoFormerV2
RoFormer升级版,主要通过结构的简化来提升速度,并通过无监督预训练和有监督预训练的结合来提升效果,从而达到了速度与效果的“双赢”。
## 介绍
- 博客:https://kexue.fm/archives/8998
## 环境
bert4keras >= 0.11.0
## 下载
- **Small版**:[chinese_roformer-v2-char_L-6_H-384_A-6.zip](https://open.zhuiyi.ai/releases/nlp/models/zhuiyi/chinese_roformer-v2-char_L-6_H-384_A-6.zip)、[百度云](https://pan.baidu.com/s/1huUrC9P60Afggo8AfiUcmA)(提取码:ttn4)
- **Base版**:[chinese_roformer-v2-char_L-12_H-768_A-12.zip](https://open.zhuiyi.ai/releases/nlp/models/zhuiyi/chinese_roformer-v2-char_L-12_H-768_A-12.zip)、[百度云](https://pan.baidu.com/s/1qcnN4LVKVe0-mnHlkN3-6Q)(提取码:pfoh)
- **Large版**:[chinese_roformer-v2-char_L-24_H-1024_A-16.zip](https://open.zhuiyi.ai/releases/nlp/models/zhuiyi/chinese_roformer-v2-char_L-24_H-1024_A-16.zip)、[百度云](https://pan.baidu.com/s/1QiJWSZrGxn8vek-8myvL6w)(提取码:npfv)## 训练
多任务训练代码参考 https://github.com/ZhuiyiTechnology/roformer-v2/tree/main/multi-task
## 配置
- **Small版**:两张3090(24G),先用无监督MLM训练了100万步(maxlen为512),然后有监督多任务训练了75万步(maxlen从64到512不等,取决于任务),batch_size为512,优化器为LAMB;
- **Base版**:四张3090(24G),先用无监督MLM训练了100万步(maxlen为512),然后有监督多任务训练了75万步(maxlen从64到512不等,取决于任务),batch_size为512,优化器为LAMB;
- **Large版**:两张A100(80G),先用无监督MLM训练了100万步(maxlen为512),然后有监督多任务训练了50万步(maxlen从64到512不等,取决于任务),batch_size为512,优化器为LAMB。
注:无监督的训练数据为280G,有监督的训练数据约为20G(77个标注数据集,构建了92个任务进行多任务训练,涵盖文本分类、文本匹配、阅读理解、信息抽取、指代消解等常见自然语言理解任务),large版的有监督训练步数更少,是因为20G的标注数据实在不够“喂饱”large级别的模型,继续训练下去出现了过拟合现象。
## 引用
Bibtex:
```tex
@techreport{roformerv2,
title={RoFormerV2: A Faster and Better RoFormer - ZhuiyiAI},
author={Jianlin Su, Shengfeng Pan, Bo Wen, Yunfeng Liu},
year={2022},
url="https://github.com/ZhuiyiTechnology/roformer-v2",
}
```## 联系
- 邮箱:[email protected]
- 追一科技:https://zhuiyi.ai