indischepartij/OpenMia-Indo-Mistral-7b-v3

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Feb 3, 2024License:cc-by-nc-4.0Architecture:Transformer0.0K Open Weights Cold

indischepartij/OpenMia-Indo-Mistral-7b-v3 is a 7 billion parameter language model developed by indischepartij, fine-tuned from Mistral-7b. This model specializes in conversational capabilities in Bahasa Indonesia. It is designed for applications requiring natural language understanding and generation in the Indonesian language, offering a context length of 4096 tokens.

Loading preview...

Overview

OpenMia-Indo-Mistral-7b-v3, developed by indischepartij, is a 7 billion parameter language model fine-tuned from the Mistral-7b architecture. Its primary focus is on enabling conversations in Bahasa Indonesia, making it a specialized tool for Indonesian language applications. The model is currently in an alpha stage of development.

Key Capabilities

  • Bahasa Indonesia Conversation: Specifically fine-tuned for engaging in dialogue and understanding prompts in Bahasa Indonesia.
  • Mistral-7b Base: Leverages the robust architecture of Mistral-7b.
  • Context Length: Supports a context window of 4096 tokens.

Performance Highlights

Evaluated on the Open LLM Leaderboard, OpenMia-Indo-Mistral-7b-v3 achieved an average score of 71.04. Notable scores include:

  • AI2 Reasoning Challenge (25-Shot): 66.13
  • HellaSwag (10-Shot): 85.47
  • MMLU (5-Shot): 64.03
  • GSM8k (5-Shot): 67.63

Prompt Format

The model utilizes a specific instruction-tuned format, expecting system and user prompts enclosed within <|im_start|> and <|im_end|> tokens, with Mia as the designated AI companion name.

Good For

  • Developing chatbots or conversational AI agents for the Indonesian market.
  • Applications requiring natural language processing in Bahasa Indonesia.
  • Experimentation with fine-tuned Mistral models for specific language tasks.