Merge branch 'main' of https://huggingface.co/wangfan/jdt-fin-roberta-wwm into main
Browse files
README.md
CHANGED
|
@@ -1,14 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
在众多业务中，越来越频繁的使用预训练语言模型（Pre-trained Language Models），为了在金融场景下各任务中取得更好效果，我们发布了jdt-fin-roberta-wwm模型
|
| 2 |
|
| 3 |
-
|
| 4 |
* `base模型`：12-layer, 768-hidden, 12-heads, 110M parameters
|
| 5 |
|
| 6 |
-
| 模型简称 |
|
| 7 |
-
|
|
| 8 |
-
| fin-roberta-wwm |
|
|
|
|
| 9 |
|
| 10 |
-
|
| 11 |
-
### 使用Huggingface-Transformers
|
| 12 |
依托于[Huggingface-Transformers](https://github.com/huggingface/transformers)，可轻松调用以上模型。
|
| 13 |
```
|
| 14 |
tokenizer = BertTokenizer.from_pretrained("MODEL_NAME")
|
|
@@ -20,4 +29,11 @@ model = BertModel.from_pretrained("MODEL_NAME")
|
|
| 20 |
| 模型名 | MODEL_NAME |
|
| 21 |
| - | - |
|
| 22 |
| fin-roberta-wwm | wangfan/jdt-fin-roberta-wwm |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
|
| 1 |
+
---
|
| 2 |
+
language: zh
|
| 3 |
+
tags:
|
| 4 |
+
- roberta-wwm
|
| 5 |
+
license: apache-2.0
|
| 6 |
+
datasets:
|
| 7 |
+
- finance
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
在众多业务中，越来越频繁的使用预训练语言模型（Pre-trained Language Models），为了在金融场景下各任务中取得更好效果，我们发布了jdt-fin-roberta-wwm模型
|
| 11 |
|
| 12 |
+
#### 模型&下载
|
| 13 |
* `base模型`：12-layer, 768-hidden, 12-heads, 110M parameters
|
| 14 |
|
| 15 |
+
| 模型简称 | 京盘下载 |
|
| 16 |
+
| :----: | :----:|
|
| 17 |
+
| fin-roberta-wwm | [Tensorflow](https://3.cn/103c-hwSS)/[Pytorch](https://3.cn/103c-izpe) |
|
| 18 |
+
| fin-roberta-wwm-large | todo |
|
| 19 |
|
| 20 |
+
#### 快速加载
|
|
|
|
| 21 |
依托于[Huggingface-Transformers](https://github.com/huggingface/transformers)，可轻松调用以上模型。
|
| 22 |
```
|
| 23 |
tokenizer = BertTokenizer.from_pretrained("MODEL_NAME")
|
| 29 |
| 模型名 | MODEL_NAME |
|
| 30 |
| - | - |
|
| 31 |
| fin-roberta-wwm | wangfan/jdt-fin-roberta-wwm |
|
| 32 |
+
| fin-roberta-wwm-large | todo |
|
| 33 |
+
|
| 34 |
+
#### 任务效果
|
| 35 |
+
| Task | NER | 关系抽取 | 事件抽取 | 指标抽取 | 实体链接 |
|
| 36 |
+
|:----:|:-- :|:------:|:-------:|:-------:|:------:|
|
| 37 |
+
| Our |93.88| 79.02 | 91.99 | 94.28| 86.72 |
|
| 38 |
+
| Roberta-wwm |93.47| 76.99 | 91.58 | 93.98| 85.20 |
|
| 39 |
|