Commit b33bb867 authored by chenzk's avatar chenzk
Browse files

Update sf.md

parent c9678312
......@@ -14,7 +14,7 @@ BERT的主模型是BERT中最重要组件,BERT通过预训练(pre-training
如图:
![image](https://developer.hpccube.com/codes/modelzoo/bert4torch/-/raw/master/doc/model.png?inline=false)
![image](https://developer.sourcefind.cn/codes/modelzoo/bert4torch/-/raw/master/doc/model.png?inline=false)
其中
......@@ -30,7 +30,7 @@ BERT模型是基于Transformer模型的,但是与原始的Transformer模型不
- 预训练(pre-training):先对大量语料进行无监督学习;
- 微调(Fine-tuning):再对少量标注语料进行监督学习,提升模型针对特定任务的表现能力。
![image](https://developer.hpccube.com/codes/modelzoo/bert4torch/-/raw/master/doc/architecture.png?inline=false)
![image](https://developer.sourcefind.cn/codes/modelzoo/bert4torch/-/raw/master/doc/architecture.png?inline=false)
## 环境配置
......@@ -61,7 +61,7 @@ python3 setup.py install
```
1.创建conda虚拟环境:
conda create -n bert4torch python=3.10
2.关于本项目DCU显卡所需的工具包、深度学习库等均可从光合开发者社区下载安装:https://developer.hpccube.com/tool/
2.关于本项目DCU显卡所需的工具包、深度学习库等均可从光合开发者社区下载安装:https://developer.sourcefind.cn/tool/
DTK驱动:dtk24.04.1
python:python3.10
torch:2.1.0
......@@ -115,7 +115,7 @@ cd examples/sequence_labeling/
./multi_train.sh
```
## result
![image](https://developer.hpccube.com/codes/modelzoo/bert4torch/-/raw/master/doc/result.png?inline=false)
![image](https://developer.sourcefind.cn/codes/modelzoo/bert4torch/-/raw/master/doc/result.png?inline=false)
### 精度
| 卡数 | 类型 | batch_size | f1 | p | r |
......@@ -140,7 +140,7 @@ cd examples/sequence_labeling/
## 源码仓库及问题反馈
- https://developer.hpccube.com/codes/modelzoo/bert4torch
- https://developer.sourcefind.cn/codes/modelzoo/bert4torch
## 参考资料
- https://github.com/Tongjilibo/bert4torch
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment