metadata

language: ko
license: mit
tags:
  - bart

Model Card for kobart-base-v2

Model Details

Model Description

BART(Bidirectional and Auto-Regressive Transformers)는 입력 텍스트 일부에 노이즈를 추가하여 이를 다시 원문으로 복구하는 autoencoder의 형태로 학습이 됩니다. 한국어 BART(이하 KoBART) 는 논문에서 사용된 Text Infilling 노이즈 함수를 사용하여 40GB 이상의 한국어 텍스트에 대해서 학습한 한국어 encoder-decoder 언어 모델입니다. 이를 통해 도출된 KoBART-base를 배포합니다.

Developed by: More information needed
Shared by [Optional]: Heewon(Haven) Jeon
Model type: Feature Extraction
Language(s) (NLP): Korean
License: MIT
Parent Model: BART
Resources for more information:
- GitHub Repo
- Model Demo Space

Uses

Direct Use

This model can be used for the task of Feature Extraction.

Downstream Use [Optional]

More information needed.

Out-of-Scope Use

The model should not be used to intentionally create hostile or alienating environments for people.

Bias, Risks, and Limitations

Significant research has explored bias and fairness issues with language models (see, e.g., Sheng et al. (2021) and Bender et al. (2021)). Predictions generated by the model may include disturbing and harmful stereotypes across protected classes; identity characteristics; and sensitive, social, and occupational groups.

Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

Training Details

Training Data

Data	# of Sentences
Korean Wiki	5M
Other corpus	0.27B

한국어 위키 백과 이외, 뉴스, 책, 모두의 말뭉치 v1.0(대화, 뉴스, ...), 청와대 국민청원 등의 다양한 데이터가 모델 학습에 사용되었습니다.

vocab 사이즈는 30,000 이며 대화에 자주 쓰이는 아래와 같은 이모티콘, 이모지 등을 추가하여 해당 토큰의 인식 능력을 올렸습니다.

😀, 😁, 😆, 😅, 🤣, .. , :-), :), -), (-:...

Training Procedure

Tokenizer

tokenizers 패키지의 Character BPE tokenizer로 학습되었습니다.

Speeds, Sizes, Times

Model	# of params	Type	# of layers	# of heads	ffn_dim	hidden_dims
`KoBART-base`	124M	Encoder	6	16	3072	768
		Decoder	6	16	3072	768

Evaluation

Testing Data, Factors & Metrics

Testing Data

More information needed

Factors

More information needed

Metrics

More information needed

Results

NSMC

acc. : 0.901

The model authors also note in the GitHub Repo:

	NSMC(acc)	KorSTS(spearman)	Question Pair(acc)
KoBART-base	90.24	81.66	94.34

Model Examination

More information needed

Environmental Impact

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

Hardware Type: More information needed
Hours used: More information needed
Cloud Provider: More information needed
Compute Region: More information needed
Carbon Emitted: More information needed

Technical Specifications [optional]

Model Architecture and Objective

More information needed

Compute Infrastructure

More information needed

Hardware

More information needed

Software

More information needed.

Citation

BibTeX:

More information needed.

Glossary [optional]

More information needed

More Information [optional]

More information needed

Model Card Authors [optional]

Heewon(Haven) Jeon in collaboration with Ezi Ozoani and the Hugging Face team

Model Card Contact

The model authors note in the GitHub Repo: KoBART 관련 이슈는 이곳에 올려주세요.

How to Get Started with the Model

Use the code below to get started with the model.

Click to expand

 from transformers import PreTrainedTokenizerFast, BartModel

tokenizer = PreTrainedTokenizerFast.from_pretrained('gogamza/kobart-base-v2')
model = BartModel.from_pretrained('gogamza/kobart-base-v2')