---
license: mit
---
This is a version of the [sealion7b](https://huggingface.co/aisingapore/sealion7b) model, sharded to 2 GB chunks.
Please refer to the previously linked repo for details on usage/implementation/etc. This model was downloaded from the original repo and is redistributed under the same license.
# SEA-LION
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 the SEA-LION 7B base model.
SEA-LION stands for Southeast Asian Languages In One Network.
## 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 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 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
### Performance Benchmarks
SEA-LION has an average performance on general tasks in English (as measured by Hugging Face's LLM Leaderboard):
| Model | ARC | HellaSwag | MMLU | TruthfulQA | Average |
|-------------|:-----:|:---------:|:-----:|:----------:|:-------:|
| SEA-LION 7B | 39.93 | 68.51 | 26.87 | 35.09 | 42.60 |
## Training Details
### Data
SEA-LION was trained on 980B tokens of the following data:
| Data Source | Unique Tokens | Multiplier | Total Tokens | Percentage |
|---------------------------|:-------------:|:----------:|:------------:|:----------:|
| RefinedWeb - English | 571.3B | 1 | 571.3B | 58.20% |
| mC4 - Chinese | 91.2B | 1 | 91.2B | 9.29% |
| mC4 - Indonesian | 3.68B | 4 | 14.7B | 1.50% |
| mC4 - Malay | 0.72B | 4 | 2.9B | 0.29% |
| mC4 - Filipino | 1.32B | 4 | 5.3B | 0.54% |
| mC4 - Burmese | 1.2B | 4 | 4.9B | 0.49% |
| mC4 - Vietnamese | 63.4B | 1 | 63.4B | 6.46% |
| mC4 - Thai | 5.8B | 2 | 11.6B | 1.18% |
| WangChanBERTa - Thai | 5B | 2 | 10B | 1.02% |
| mC4 - Lao | 0.27B | 4 | 1.1B | 0.12% |
| mC4 - Khmer | 0.97B | 4 | 3.9B | 0.40% |
| mC4 - Tamil | 2.55B | 4 | 10.2B | 1.04% |
| the Stack - Python | 20.9B | 2 | 41.8B | 4.26% |
| the Stack - Javascript | 55.6B | 1 | 55.6B | 5.66% |
| the Stack - Shell | 1.2B5 | 2 | 2.5B | 0.26% |
| the Stack - SQL | 6.4B | 2 | 12.8B | 1.31% |
| the Stack - Markdown | 26.6B | 1 | 26.6B | 2.71% |
| RedPajama - StackExchange | 21.2B | 1 | 21.2B | 2.16% |
| RedPajama - ArXiv | 30.6B | 1 | 30.6B | 3.12% |
### Infrastructure
SEA-LION was trained using [MosaicML Composer](https://github.com/mosaicml/composer)
on the following hardware:
| Training Details | SEA-LION 7B |
|----------------------|:------------:|
| AWS EC2 p4d.24xlarge | 32 instances |
| Nvidia A100 40GB GPU | 256 |
| Training Duration | 22 days |
### Configuration
| HyperParameter | SEA-LION 7B |
|-------------------|:------------------:|
| Precision | bfloat16 |
| Optimizer | decoupled_adamw |
| Scheduler | cosine_with_warmup |
| Learning Rate | 6.0e-5 |
| Global Batch Size | 2048 |
| Micro Batch Size | 4 |
## Technical Specifications
### Model Architecture and Objective
SEA-LION is a decoder model using the MPT architecture.
| Parameter | SEA-LION 7B |
|-----------------|:-----------:|
| Layers | 32 |
| d_model | 4096 |
| head_dim | 32 |
| Vocabulary | 256000 |
| Sequence Length | 2048 |
### Tokenizer Details
We sample 20M lines from the training data to train the tokenizer.
The framework for training is [SentencePiece](https://github.com/google/sentencepiece).
The tokenizer type is Byte-Pair Encoding (BPE).
## The Team
Lam Wen Zhi Clarence
Leong Wei Qi
Li Yier
Liu Bing Jie Darius
Lovenia Holy
Montalan Jann Railey
Ng Boon Cheong Raymond
Ngui Jian Gang
Nguyen Thanh Ngan
Ong Tat-Wee David
Rengarajan Hamsawardhini
Susanto Yosephine
Tai Ngee Chia
Tan Choon Meng
Teo Jin Howe
Teo Eng Sipp Leslie
Teo Wei Yi
Tjhi William
Yeo Yeow Tong
Yong Xianbin
## 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](https://forms.gle/sLCUVb95wmGf43hi6)
[Link to SEA-LION's GitHub repository](https://github.com/aisingapore/sealion)
## 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
```bibtex
@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}
}
```