SEA-LION-v1-3B

SEA-LION-v1-3B

SEA-LION is a collection of Large Language Models (LLMs) which has been pretrained and instruct-tuned for the Southeast Asia (SEA) region. The size of the models range from 3 billion to 7 billion parameters. This is the card for SEA-LION-v1-3B.

SEA-LION stands for <i>Southeast Asian Languages In One Network</i>.

Model Details

Model Description

The SEA-LION model is a significant leap forward in the field of Natural Language Processing, specifically trained to understand the SEA regional context.

SEA-LION-v1-3B is built on the robust MPT architecture and has a vocabulary size of 256K.

For tokenization, the model employs our custom SEABPETokenizer, which is specially tailored for SEA languages, ensuring optimal model performance.

The training data for SEA-LION-v1-3B encompasses 980B tokens.

• Developed by: Products Pillar, AI Singapore
• Funded by: Singapore NRF
• Model type: Decoder
• Languages: English, Chinese, Indonesian, Malay, Thai, Vietnamese, Filipino, Tamil, Burmese, Khmer, Lao
• License: MIT License

Lineage & Versioning

The training counts and dataset mixture details reported in this model card reflect the exact constructed training pool consumed during this specific model run. Figures may differ slightly from public dataset releases, which represent downloadable open-source subsets of the broader corpus.

Performance Benchmarks

SEA-LION-v1-3B has an average performance on general tasks in English (as measured by Hugging Face's LLM Leaderboard):

Training Details

Data

SEA-LION-v1-3B was trained on 980B tokens of the following data:

Infrastructure

SEA-LION-v1-3B was trained using MosaicML Composer on the following hardware:

Configuration

Technical Specifications

Model Architecture and Objective

SEA-LION-v1-3B is a decoder model using the MPT architecture.

Tokenizer Details

We sample 20M lines from the training data to train the tokenizer.<br> The framework for training is SentencePiece.<br> The tokenizer type is Byte-Pair Encoding (BPE).

The Team

Lam Wen Zhi Clarence<br> Leong Wei Qi<br> Li Yier<br> Liu Bing Jie Darius<br> Lovenia Holy<br> Montalan Jann Railey<br> Ng Boon Cheong Raymond<br> Ngui Jian Gang<br> Nguyen Thanh Ngan<br> Ong Tat-Wee David<br> Rengarajan Hamsawardhini<br> Susanto Yosephine<br> Tai Ngee Chia<br> Tan Choon Meng<br> Teo Jin Howe<br> Teo Eng Sipp Leslie<br> Teo Wei Yi<br> Tjhi William<br> Yeo Yeow Tong<br> Yong Xianbin<br>

Acknowledgements

AI Singapore is a national programme supported by the National Research Foundation, Singapore and hosted by the National University of Singapore. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not reflect the views of National Research Foundation, Singapore.

Contact

For more info, please contact us using this SEA-LION Inquiry Form

Link to SEA-LION's GitHub repository

Disclaimer

This the repository for the base model. The model has not been aligned for safety. Developers and users should perform their own safety fine-tuning and related security measures. In no event shall the authors be held liable for any claim, damages, or other liability arising from the use of the released weights and codes.

References

Thai Pre-Training Data Reference

@misc{lowphansirikul2021wangchanberta,
    title={WangchanBERTa: Pretraining transformer-based Thai Language Models},
    author={Lalita Lowphansirikul and Charin Polpanumas and Nawat Jantrakulchai and Sarana Nutanong},
    year={2021},
    eprint={2101.09635},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}