OpenBuddy/openbuddy-llama3-8b-v21.2-32k Overview

OpenBuddy/openbuddy-llama3-8b-v21.2-32k is an 8 billion parameter instruction-tuned language model developed by the OpenBuddy team. It is built upon the Meta Llama-3-8B-Instruct base model and is designed to function as a helpful, respectful, and honest AI assistant. The model supports a context window of 8192 tokens.

Key Capabilities

• Multilingual Support: Primarily intended for use in English and Chinese.
• Instruction Following: Fine-tuned for conversational interactions, adhering to a specific prompt format for optimal performance.
• Safety and Ethics: Designed to produce responses that are socially unbiased, positive, and free from harmful, political, religious, unethical, racist, sexist, toxic, dangerous, or illegal content.
• OpenAI API Compatibility: The defined prompt format in tokenizer_config.json allows for direct deployment with vllm to create an OpenAI-like API service.

Intended Use Cases

This model is suitable for general-purpose chatbot applications where a focus on multilingual (English and Chinese) and ethical responses is desired. It is designed for interactive conversations and can be integrated into various AI assistant roles. Users should be aware of the inherent limitations common to all LLMs, including the potential for erroneous or undesirable outputs, and should not use it in critical or high-stakes scenarios.