Joelzhang commited on
Commit
20041e9
1 Parent(s): ad07ee1

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +54 -13
README.md CHANGED
@@ -14,11 +14,41 @@ widget:
14
  - text: "今天心情不好[SEP]今天很开心"
15
 
16
  ---
17
- # Erlangshen-MegatronBert-1.3B-Similarity, model (Chinese),one model of [Fengshenbang-LM](https://github.com/IDEA-CCNL/Fengshenbang-LM).
18
- We collect 20 paraphrace datasets in the Chinese domain for finetune, with a total of 2773880 samples. Our model is mainly based on [MegatronBert-1.3B](https://huggingface.co/IDEA-CCNL/Erlangshen-MegatronBert-1.3B)
19
 
20
- ## Usage
21
- ```python
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
22
  from transformers import AutoModelForSequenceClassification
23
  from transformers import BertTokenizer
24
  import torch
@@ -31,18 +61,29 @@ textb='今天心情不好'
31
 
32
  output=model(torch.tensor([tokenizer.encode(texta,textb)]))
33
  print(torch.nn.functional.softmax(output.logits,dim=-1))
34
-
35
  ```
36
- ## Scores on downstream chinese tasks(The dev datasets of BUSTM and AFQMC may exist in the train set)
37
- | Model | BQ | BUSTM | AFQMC |
38
- | :--------: | :-----: | :----: | :-----: |
39
- | Erlangshen-Roberta-110M-Similarity | 85.41 | 95.18 | 81.72 |
40
- | Erlangshen-Roberta-330M-Similarity | 86.21 | 99.29 | 93.89 |
41
- | Erlangshen-MegatronBert-1.3B-Similarity | 86.31 | - | - |
42
 
43
- ## Citation
44
- If you find the resource is useful, please cite the following website in your paper.
 
 
 
 
 
 
 
 
 
 
 
 
45
  ```
 
 
 
 
 
 
46
  @misc{Fengshenbang-LM,
47
  title={Fengshenbang-LM},
48
  author={IDEA-CCNL},
 
14
  - text: "今天心情不好[SEP]今天很开心"
15
 
16
  ---
17
+ # Erlangshen-MegatronBert-1.3B-Similarity
 
18
 
19
+ - Github: [Fengshenbang-LM](https://github.com/IDEA-CCNL/Fengshenbang-LM)
20
+ - Docs: [Fengshenbang-Docs](https://fengshenbang-doc.readthedocs.io/)
21
+
22
+ ## 简介 Brief Introduction
23
+
24
+ 2021年登顶FewCLUE和ZeroCLUE的中文BERT,在数个改写任务微调后的版本
25
+
26
+ This is the fine-tuned version of the Chinese BERT model on several semantic matching and similarity datasets, which topped FewCLUE and ZeroCLUE benchmark in 2021
27
+
28
+ ## 模型分类 Model Taxonomy
29
+
30
+ | 需求 Demand | 任务 Task | 系列 Series | 模型 Model | 参数 Parameter | 额外 Extra |
31
+ | :----: | :----: | :----: | :----: | :----: | :----: |
32
+ | 通用 General | 自然语言理解 NLU | 二郎神 Erlangshen | MegatronBert | 1.3B | Similarity |
33
+
34
+ ## 模型信息 Model Information
35
+ We collect 20 paraphrace datasets in the Chinese domain for finetune, with a total of 2773880 samples
36
+ 基于[Erlangshen-MegatronBert-1.3B](https://huggingface.co/IDEA-CCNL/Erlangshen-MegatronBert-1.3B),我们在收集的20个用于finetune的中文领域的改写数据集,总计227347个样本上微调了一个Similarity版本。
37
+
38
+ Based on [Erlangshen-MegatronBert-1.3B] (https://huggingface.co/IDEA-CCNL/Erlangshen-MegatronBert-1.3B), we fine-tuned a similarity version on 8 Chinese paraphrace datasets, with totaling 227,347 samples.
39
+
40
+ ### 下游效果 Performance
41
+
42
+ | Model | BQ | BUSTM | AFQMC |
43
+ | :--------: | :-----: | :----: | :-----: |
44
+ | Erlangshen-Roberta-110M-Similarity | 85.41 | 95.18 | 81.72 |
45
+ | Erlangshen-Roberta-330M-Similarity | 86.21 | 99.29 | 93.89 |
46
+ | Erlangshen-MegatronBert-1.3B-Similarity | 86.31 | - | - |
47
+
48
+
49
+ ## 使用 Usage
50
+
51
+ ``` python
52
  from transformers import AutoModelForSequenceClassification
53
  from transformers import BertTokenizer
54
  import torch
 
61
 
62
  output=model(torch.tensor([tokenizer.encode(texta,textb)]))
63
  print(torch.nn.functional.softmax(output.logits,dim=-1))
 
64
  ```
 
 
 
 
 
 
65
 
66
+ ## 引用 Citation
67
+
68
+ 如果您在您的工作中使用了我们的模型,可以引用我们的[论文](https://arxiv.org/abs/2209.02970):
69
+
70
+ If you are using the resource for your work, please cite the our [paper](https://arxiv.org/abs/2209.02970):
71
+
72
+ ```text
73
+ @article{fengshenbang,
74
+ author = {Junjie Wang and Yuxiang Zhang and Lin Zhang and Ping Yang and Xinyu Gao and Ziwei Wu and Xiaoqun Dong and Junqing He and Jianheng Zhuo and Qi Yang and Yongfeng Huang and Xiayu Li and Yanghan Wu and Junyu Lu and Xinyu Zhu and Weifeng Chen and Ting Han and Kunhao Pan and Rui Wang and Hao Wang and Xiaojun Wu and Zhongshen Zeng and Chongpei Chen and Ruyi Gan and Jiaxing Zhang},
75
+ title = {Fengshenbang 1.0: Being the Foundation of Chinese Cognitive Intelligence},
76
+ journal = {CoRR},
77
+ volume = {abs/2209.02970},
78
+ year = {2022}
79
+ }
80
  ```
81
+
82
+ 也可以引用我们的[网站](https://github.com/IDEA-CCNL/Fengshenbang-LM/):
83
+
84
+ You can also cite our [website](https://github.com/IDEA-CCNL/Fengshenbang-LM/):
85
+
86
+ ```text
87
  @misc{Fengshenbang-LM,
88
  title={Fengshenbang-LM},
89
  author={IDEA-CCNL},