aisingapore/Llama-SEA-LION-v2-8B-IT

TEXT GENERATIONConcurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:8kTool Calling:SupportedPublished:Jul 30, 2024License:llama3Architecture:Transformer0.0K Featherless Exclusive Cold

Llama-SEA-LION-v2-8B-IT is an 8 billion parameter decoder-only Large Language Model developed by AI Singapore, built on the Llama 3 architecture with an 8192 token context length. It is instruction-tuned for Southeast Asian languages, specifically English, Indonesian, Thai, and Vietnamese. This model excels in general language capabilities and instruction-following tasks across these languages, making it suitable for applications requiring multilingual understanding and generation in the SEA region.

Loading preview...

Llama-SEA-LION-v2-8B-IT: Southeast Asian Language Instruction-Tuned Model

Developed by AI Singapore, Llama-SEA-LION-v2-8B-IT is an 8 billion parameter instruction-tuned model based on the Llama 3 decoder architecture. It is specifically designed and optimized for the Southeast Asian (SEA) region, supporting English, Indonesian, Thai, and Vietnamese with a context length of 8192 tokens.

Key Capabilities & Features

  • Multilingual Support: Instruction-tuned in English and key ASEAN languages (Indonesian, Thai, Vietnamese).
  • Llama 3 Architecture: Utilizes the robust Llama 3 8B Instruct tokenizer and underlying architecture.
  • Comprehensive Evaluation: Benchmarked on general language capabilities using SEA-HELM (BHASA) across tasks like QA, Sentiment Analysis, Translation, and Summarization. Instruction-following is evaluated with SEA-IFEval and SEA-MTBench, adapted for target languages.
  • Regional Focus: Part of the SEA-LION (Southeast Asian Languages In One Network) initiative, aiming to enhance LLM performance for the region.

When to Use This Model

  • Multilingual Applications: Ideal for use cases requiring understanding and generation in English, Indonesian, Thai, or Vietnamese.
  • Instruction Following: Suited for tasks where adherence to specific instructions and multi-turn conversational abilities are critical.
  • Regional NLP: Developers targeting the Southeast Asian market can leverage its specialized training for improved relevance and performance.

Important Considerations

  • Safety Alignment: The model has not been aligned for safety; users are advised to perform their own safety fine-tuning.
  • Limitations: Like many LLMs, it may hallucinate or generate irrelevant content and can exhibit inconsistencies in reasoning.

Finetunes
3 models