Llama-SEA-LION-v3-70B-IT

SEA-LION is a collection of Large Language Models (LLMs) which have been pretrained and instruct-tuned for the Southeast Asia (SEA) region.

Llama-SEA-LION-v3-70B-IT is a multilingual model that has been fine-tuned in two stages on approximately 12.3M English instruction-completion pairs alongside a pool of 4.5M Southeast Asian instruction-completion pairs from SEA languages such as Indonesian, Javanese, Sundanese, Tamil, Thai, and Vietnamese.

SEA-LION stands for Southeast Asian Languages In One Network.

• Developed by: Products Pillar, AI Singapore
• Funded by: Singapore NRF
• Model type: Decoder
• Languages supported: Burmese, Chinese, English, Filipino, Indonesia, Javanese, Khmer, Lao, Malay, Sundanese, Tamil, Thai, Vietnamese
• License: Llama 3.1 Community License

Description

This repo contains GGUF format model files for aisingapore/Llama-SEA-LION-v3-70B-IT.

Model Weights Included in this repository:

• Llama-SEA-LION-v3-70B-IT-F16
• Llama-SEA-LION-v3-70B-IT-Q2_K
• Llama-SEA-LION-v3-70B-IT-Q3_K_M
• Llama-SEA-LION-v3-70B-IT-Q4_0
• Llama-SEA-LION-v3-70B-IT-Q4_K_M
• Llama-SEA-LION-v3-70B-IT-Q5_0
• Llama-SEA-LION-v3-70B-IT-Q5_K_M
• Llama-SEA-LION-v3-70B-IT-Q6_K
• Llama-SEA-LION-v3-70B-IT-Q8_0

[!NOTE] Take note that some GGUFs are split into parts. Most tools such as llama.cpp and those built on it do support split GGUFs, pointing the platform to the first split will be sufficient for it to function. In the event where a merge is necessary, it can be done using llama.cpp's gguf-split: ./gguf-split --merge ./path/to/first-split ./path/to/output-gguf More details: gguf-split guide & README

Caveats

It is important for users to be aware that our model exhibits certain limitations that warrant consideration. Like many LLMs, the model can hallucinate and occasionally generates irrelevant content, introducing fictional elements that are not grounded in the provided context. Users should also exercise caution in interpreting and validating the model's responses due to the potential inconsistencies in its reasoning.

Limitations

Safety

Current SEA-LION models, including this commercially permissive release, have 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.

Technical Specifications

Fine-Tuning Details

Llama-SEA-LION-v3-70B-IT was tuned using a combination of a full parameter fine-tune, on-policy alignment, and model merges of the best performing checkpoints. The training process for fine-tuning was approximately 3200 GPU hours, on a single node of 8x H100-80GB GPUs.

Data

Llama-SEA-LION-v3-70B-IT was trained on a wide range of synthetic instructions, alongside publicly available instructions hand-curated by the team with the assistance of native speakers. In addition, special care was taken to ensure that the datasets used had commercially permissive licenses through verification with the original data source.

Lineage & Versioning

This model card serves as an immutable record of the final released checkpoint. Consequently, the hardware specifications, compute hours, and precise dataset volumes (such as final filtered instruction counts) reported here reflect the exact production run used to generate this specific artifact. These figures may differ from the aggregate totals, pre-filtered data pools, or preliminary experimental runs (e.g., initial H100 benchmarks) documented in our accompanying research papers.

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.

Call for Contributions

We encourage researchers, developers, and language enthusiasts to actively contribute to the enhancement and expansion of SEA-LION. Contributions can involve identifying and reporting bugs, sharing pre-training, instruction, and preference data, improving documentation usability, proposing and implementing new model evaluation tasks and metrics, or training versions of the model in additional Southeast Asian languages. Join us in shaping the future of SEA-LION by sharing your expertise and insights to make these models more accessible, accurate, and versatile. Please check out our GitHub for further information on the call for contributions.

The Team

Chan Adwin, Cheng Nicholas, Choa Esther, Huang Yuli, Lau Wayne, Lee Chwan Ren, Leong Wai Yi, Leong Wei Qi, Limkonchotiwat Peerat, Liu Bing Jie Darius, Montalan Jann Railey, Ng Boon Cheong Raymond, Ngui Jian Gang, Nguyen Thanh Ngan, Ong Brandon, Ong Tat-Wee David, Ong Zhi Hao, Rengarajan Hamsawardhini, Siow Bryan, Susanto Yosephine, Tai Ngee Chia, Tan Choon Meng, Teng Walter, Teo Eng Sipp Leslie, Teo Wei Yi, Tjhi William, Venkatadri Hulagadri Adithya, 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 the National Research Foundation or the National University of Singapore.

Contact

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

Link to SEA-LION's GitHub repository

Disclaimer

This is the repository for the commercial instruction-tuned 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 claims, damages, or other liabilities arising from the use of the released weights and codes.