nazneen commited on
Commit
a186ef5
1 Parent(s): d13dd65

model documentation

Browse files
Files changed (1) hide show
  1. README.md +186 -31
README.md CHANGED
@@ -1,43 +1,198 @@
 
1
  ---
 
 
 
2
  license: apache-2.0
3
- language: [en, ko]
4
  tags:
5
- - t5
6
- eos_token: "</s>"
7
  widget:
8
- - text: 아버지가 방에 들어가신다.</s>
9
  ---
10
 
11
- # ke-t5 base
12
-
13
- Pretrained T5 Model on Korean and English. See [Github](https://github.com/AIRC-KETI/ke-t5) and [Paper](https://aclanthology.org/2021.findings-emnlp.33/) [Korean paper](https://koreascience.kr/article/CFKO202130060717834.pdf) for more details.
14
-
15
- ## How to use
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
16
 
17
  ```python
18
- from transformers import AutoModel, AutoTokenizer
19
 
20
- model = AutoModel.from_pretrained("KETI-AIR/ke-t5-base")
21
  tokenizer = AutoTokenizer.from_pretrained("KETI-AIR/ke-t5-base")
22
- ```
23
 
24
- ## BibTeX entry and citation info
 
 
 
 
 
25
 
26
- ```bibtex
27
- @inproceedings{kim-etal-2021-model-cross,
28
- title = "A Model of Cross-Lingual Knowledge-Grounded Response Generation for Open-Domain Dialogue Systems",
29
- author = "Kim, San and
30
- Jang, Jin Yea and
31
- Jung, Minyoung and
32
- Shin, Saim",
33
- booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2021",
34
- month = nov,
35
- year = "2021",
36
- address = "Punta Cana, Dominican Republic",
37
- publisher = "Association for Computational Linguistics",
38
- url = "https://aclanthology.org/2021.findings-emnlp.33",
39
- doi = "10.18653/v1/2021.findings-emnlp.33",
40
- pages = "352--365",
41
- abstract = "Research on open-domain dialogue systems that allow free topics is challenging in the field of natural language processing (NLP). The performance of the dialogue system has been improved recently by the method utilizing dialogue-related knowledge; however, non-English dialogue systems suffer from reproducing the performance of English dialogue systems because securing knowledge in the same language with the dialogue system is relatively difficult. Through experiments with a Korean dialogue system, this paper proves that the performance of a non-English dialogue system can be improved by utilizing English knowledge, highlighting the system uses cross-lingual knowledge. For the experiments, we 1) constructed a Korean version of the Wizard of Wikipedia dataset, 2) built Korean-English T5 (KE-T5), a language model pre-trained with Korean and English corpus, and 3) developed a knowledge-grounded Korean dialogue model based on KE-T5. We observed the performance improvement in the open-domain Korean dialogue model even only English knowledge was given. The experimental results showed that the knowledge inherent in cross-lingual language models can be helpful for generating responses in open dialogue systems.",
42
- }
43
- ```
 
1
+
2
  ---
3
+ language:
4
+ - en
5
+ - ko
6
  license: apache-2.0
 
7
  tags:
8
+ - t5
9
+ eos_token: </s>
10
  widget:
11
+ - text: 아버지가 방에 들어가신다.</s>
12
  ---
13
 
14
+ # Model Card for ke-t5-base
15
+
16
+
17
+ # Model Details
18
+
19
+ ## Model Description
20
+
21
+ The developers of the Text-To-Text Transfer Transformer (T5) [write](https://ai.googleblog.com/2020/02/exploring-transfer-learning-with-t5.html):
22
+
23
+ > With T5, we propose reframing all NLP tasks into a unified text-to-text-format where the input and output are always text strings, in contrast to BERT-style models that can only output either a class label or a span of the input. Our text-to-text framework allows us to use the same model, loss function, and hyperparameters on any NLP task.
24
+
25
+ T5-Base is the checkpoint with 220 million parameters.
26
+
27
+
28
+ - **Developed by:** Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, Peter J. Liu.
29
+ - **Shared by [Optional]:** Korea Electronics Technology Institute Artificial Intelligence Research Center
30
+ - **Model type:** Text Generation
31
+ - **Language(s) (NLP):**More information needed
32
+ - **License:** More information needed
33
+ - **Related Models:**
34
+ - **Parent Model:** T5
35
+ - **Resources for more information:**
36
+ - [GitHub Repo](https://github.com/google-research/text-to-text-transfer-transformer#released-model-checkpoints)
37
+ - [Associated Paper](https://jmlr.org/papers/volume21/20-074/20-074.pdf)
38
+ - [Blog Post](https://ai.googleblog.com/2020/02/exploring-transfer-learning-with-t5.html)
39
+
40
+ # Uses
41
+
42
+
43
+ ## Direct Use
44
+
45
+ The developers write in a [blog post](https://ai.googleblog.com/2020/02/exploring-transfer-learning-with-t5.html) that the model:
46
+
47
+ > Our text-to-text framework allows us to use the same model, loss function, and hyperparameters on any NLP task, including machine translation, document summarization, question answering, and classification tasks (e.g., sentiment analysis). We can even apply T5 to regression tasks by training it to predict the string representation of a number instead of the number itself
48
+
49
+
50
+ ## Downstream Use [Optional]
51
+
52
+ More information needed
53
+
54
+ ## Out-of-Scope Use
55
+
56
+ The model should not be used to intentionally create hostile or alienating environments for people.
57
+
58
+ # Bias, Risks, and Limitations
59
+
60
+ Significant research has explored bias and fairness issues with language models (see, e.g., [Sheng et al. (2021)](https://aclanthology.org/2021.acl-long.330.pdf) and [Bender et al. (2021)](https://dl.acm.org/doi/pdf/10.1145/3442188.3445922)). Predictions generated by the model may include disturbing and harmful stereotypes across protected classes; identity characteristics; and sensitive, social, and occupational groups.
61
+
62
+
63
+ ## Recommendations
64
+
65
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
66
+
67
+
68
+ # Training Details
69
+
70
+ ## Training Data
71
+
72
+ The model is pre-trained on the [Colossal Clean Crawled Corpus (C4)](https://www.tensorflow.org/datasets/catalog/c4), which was developed and released in the context of the same [research paper](https://jmlr.org/papers/volume21/20-074/20-074.pdf) as T5.
73
+
74
+ The model was pre-trained on a on a **multi-task mixture of unsupervised (1.) and supervised tasks (2.)**.
75
+
76
+ See the [t5-base model card](https://huggingface.co/t5-base?text=My+name+is+Wolfgang+and+I+live+in+Berlin) for further information.
77
+
78
+ ## Training Procedure
79
+
80
+
81
+ ### Preprocessing
82
+
83
+ More information needed
84
+
85
+ ### Speeds, Sizes, Times
86
+
87
+ More information needed
88
+
89
+ # Evaluation
90
+
91
+
92
+ ## Testing Data, Factors & Metrics
93
+
94
+ ### Testing Data
95
+
96
+ The developers evaluated the model on 24 tasks, see the [research paper](https://jmlr.org/papers/volume21/20-074/20-074.pdf) for full details.
97
+
98
+ ### Factors
99
+ More information needed
100
+
101
+
102
+ ### Metrics
103
+
104
+ More information needed
105
+ ## Results
106
+
107
+ For full results for T5-Base, see the [research paper](https://jmlr.org/papers/volume21/20-074/20-074.pdf), Table 14.
108
+
109
+ # Model Examination
110
+
111
+ More information needed
112
+
113
+ # Environmental Impact
114
+
115
+
116
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
117
+
118
+ - **Hardware Type:** Google Cloud TPU Pods
119
+ - **Hours used:** More information needed
120
+ - **Cloud Provider:** GCP
121
+ - **Compute Region:** More information needed
122
+ - **Carbon Emitted:** More information needed
123
+
124
+ # Technical Specifications [optional]
125
+
126
+ ## Model Architecture and Objective
127
+
128
+ More information needed
129
+
130
+ ## Compute Infrastructure
131
+
132
+ More information needed
133
+
134
+ ### Hardware
135
+
136
+ More information needed
137
+
138
+ ### Software
139
+ More information needed
140
+
141
+ # Citation
142
+
143
+
144
+ **BibTeX:**
145
+ ```bibtex
146
+ @article{2020t5,
147
+ author = {Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu},
148
+ title = {Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer},
149
+ journal = {Journal of Machine Learning Research},
150
+ year = {2020},
151
+ volume = {21},
152
+ number = {140},
153
+ pages = {1-67},
154
+ url = {http://jmlr.org/papers/v21/20-074.html}
155
+ }
156
+
157
+ ```
158
+
159
+ **APA:**
160
+ - Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., ... & Liu, P. J. (2020). Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res., 21(140), 1-67.
161
+
162
+
163
+
164
+ # Glossary [optional]
165
+ More information needed
166
+
167
+ # More Information [optional]
168
+
169
+ More information needed
170
+
171
+ # Model Card Authors [optional]
172
+
173
+
174
+ Korea Electronics Technology Institute Artificial Intelligence Research Center in collaboration with Ezi Ozoani and the Hugging Face team
175
+
176
+ # Model Card Contact
177
+
178
+ More information needed
179
+
180
+ # How to Get Started with the Model
181
+
182
+ Use the code below to get started with the model.
183
+
184
+ <details>
185
+ <summary> Click to expand </summary>
186
 
187
  ```python
188
+ from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
189
 
 
190
  tokenizer = AutoTokenizer.from_pretrained("KETI-AIR/ke-t5-base")
 
191
 
192
+ model = AutoModelForSeq2SeqLM.from_pretrained("KETI-AIR/ke-t5-base")
193
+ ```
194
+
195
+ See the [Hugging Face T5](https://huggingface.co/docs/transformers/model_doc/t5#transformers.T5Model) docs and a [Colab Notebook](https://colab.research.google.com/github/google-research/text-to-text-transfer-transformer/blob/main/notebooks/t5-trivia.ipynb) created by the model developers for more examples.
196
+ </details>
197
+
198