aisingapore/llama3-8b-cpt-sea-lionv2.1-instruct-gguf

Llama 3 8B CPT SEA-Lionv2.1 Instruct

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

Llama 3 8B CPT SEA-Lionv2.1 Instruct is a multilingual model which has been fine-tuned with around 100,000 English instruction-completion pairs alongside a smaller pool of around 50,000 instruction-completion pairs from other ASEAN languages, such as Indonesian, Thai and Vietnamese. These instructions have been carefully curated and rewritten to ensure the model was trained on truly open, commercially permissive and high quality datasets.

Llama3 8B CPT SEA-Lionv2.1 Instruct has undergone additional supervised fine-tuning and alignment compared to the now deprecated Llama3 8B CPT SEA-Lionv2 Instruct. These improvements have increased the model's capabilities in chat interactions and its ability to follow instructions accurately.

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

• Developed by: Products Pillar, AI Singapore
• Funded by: Singapore NRF
• Model type: Decoder
• Languages: English, Indonesian, Thai, Vietnamese, Tamil
• License: Llama 3 Community License

Description

This repo contains GGUF format model files for aisingapore/llama3-8b-cpt-sea-lionv2.1-instruct.

Model Weights Included in this repository:

• llama3-8b-cpt-sea-lionv2.1-instruct-Q2_K
• llama3-8b-cpt-sea-lionv2.1-instruct-Q3_K_M
• llama3-8b-cpt-sea-lionv2.1-instruct-Q4_0
• llama3-8b-cpt-sea-lionv2.1-instruct-Q4_K_M
• llama3-8b-cpt-sea-lionv2.1-instruct-Q5_0
• llama3-8b-cpt-sea-lionv2.1-instruct-Q5_K_M
• llama3-8b-cpt-sea-lionv2.1-instruct-Q6_K
• llama3-8b-cpt-sea-lionv2.1-instruct-Q8_0

Usage

Llama 3 8B CPT SEA-Lionv2.1 Instruct GGUF files have been tested on the latest version of llama.cpp, from pull request #8772

Prompt Template:

<|begin_of_text|><|start_header_id|>system<|end_header_id|>

{{system_prompt}}<|eot_id|>
<|start_header_id|>user<|end_header_id|>

{{prompt}}<|eot_id|>
<|start_header_id|>assistant<|end_header_id|>

Recommended llama.cpp command:

To execute the following commands, ensure you are in the llama.cpp root directory and that your models are located in the models folder:

# Running one-time input prompt
./llama-cli -m models/llama3-8b-cpt-sea-lionv2.1-instruct/llama3-8B-cpt-sea-lionv2.1-instruct-Q4_K_M.gguf -ngl -1 --temp 0 -n 128 -p "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nYou are a helpful assistant who answers succinctly.<|eot_id|>\n<|start_header_id|>user<|end_header_id|>\n\nWhat is a sea lion?<|eot_id|>\n<|start_header_id|>assistant<|end_header_id|>\n\n"

# Running in conversation mode
./llama-cli -m models/llama3-8b-cpt-sea-lionv2.1-instruct/llama3-8B-cpt-sea-lionv2.1-instruct-Q4_K_M.gguf -ngl -1 --temp 0 -n 128 -p "You are a helpful assistant who answers succinctly." --color -cnv --chat-template llama3

Please refer to the llama.cpp documentation for adjusting the parameters.

To convert & quantize your own SEA-LION model:

Given that you are in the llama.cpp root directory:

python convert-hf-to-gguf.py {{model path}}
./quantize ggml-model-f16.gguf {{Quant Type}}

For more detailed instructions on conversion and quantization, please refer to llama.cpp documentation.

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

The Llama 3 8B CPT SEA-Lionv2.1 Instruct was fine-tuned using 8x A100-40GB using parameter efficient fine tuning in the form of LoRA.

Data

Llama 3 8B CPT SEA-Lionv2.1 Instruct was trained on a wide range of instructions that were manually and stringently verified by our team. A large portion of the effort was dedicated to ensuring that each instruction-completion pair that the model sees is of a high quality and any errors were corrected and rewritten by native speakers or else dropped from our mix.

In addition, special care was taken to ensure that the datasets used had commercially permissive licenses through verification with the original data source.

Link to dataset: coming soon

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

Choa Esther
Cheng Nicholas
Huang Yuli
Lau Wayne
Lee Chwan Ren
Leong Wai Yi
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 Brandon
Ong Tat-Wee David
Ong Zhi Hao
Rengarajan Hamsawardhini
Siow Bryan
Susanto Yosephine
Tai Ngee Chia
Tan Choon Meng
Teo Eng Sipp Leslie
Teo Wei Yi
Tjhi William
Teng Walter
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.