kalisai/Nusantara-1.8b-Indo-Chat
Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:1.8BQuant:BF16Ctx Length:32kPublished:Mar 5, 2024License:apache-2.0Architecture:Transformer Open Weights Warm

The Nusantara-1.8b-Indo-Chat model by Kalis AI is a 1.8 billion parameter, transformer-based decoder-only language model, fine-tuned from Qwen1.5 for Bahasa Indonesia. It is designed for chat-based applications, excelling in Question-Answering and instruction following in Indonesian. With a context length of 32768 tokens, it serves as a versatile open-weight model for Indonesian language processing.

Loading preview...

Nusantara-1.8b-Indo-Chat Overview

Nusantara-1.8b-Indo-Chat is a 1.8 billion parameter language model developed by Kalis AI, specifically fine-tuned for Bahasa Indonesia. It is built upon the Qwen1.5 architecture and leverages domain-specific datasets to enhance its performance in Indonesian language tasks. The model is designed to function as a chat-implemented language model, capable of handling Question-Answering and responding to instructions given in Bahasa Indonesia.

Key Capabilities

  • Bahasa Indonesia Proficiency: Optimized for understanding and generating text in Bahasa Indonesia.
  • Chat-based Interaction: Excels in conversational AI, including question-answering and instruction following.
  • Multilingual Support: While primarily focused on Indonesian, it also supports English.
  • Qwen1.5 Foundation: Benefits from the robust architecture of the Qwen1.5 language model series.

Important Considerations

  • Hallucination Risk: As a model with less than 4 billion parameters, it may be prone to hallucination. Users should be aware of this limitation.
  • Uncensored Training: The model was trained with uncensored datasets, which may lead to negative impacts. Users are responsible for the outcomes of its usage.
  • License: Licensed under Apache-2.0, but usage must also comply with the Qwen License.

Good For

  • Developing AI assistants and chatbots for the Indonesian market.
  • Applications requiring instruction-following and Q&A in Bahasa Indonesia.
  • Research and development in Indonesian natural language processing.