Overview

rinna/youri-7b-chat is an instruction-tuned version of the rinna/youri-7b model, designed for chat-style input and responses. Developed by rinna, this 7 billion parameter model utilizes a 32-layer, 4096-hidden-size transformer architecture, consistent with the Llama 2 framework.

Key Capabilities

• Bilingual Instruction Following: Fine-tuned on a comprehensive mix of English and Japanese datasets, enabling robust performance in both languages.
• Chat-Style Interaction: Optimized for conversational AI, capable of understanding and generating responses in a dialogue format.
• Diverse Training Data: Leverages datasets such as Databricks Dolly, Anthropic HH RLHF, FLAN, and specific Japanese LLM datasets to enhance its instruction-following abilities and general knowledge.

Good For

• Multilingual Chatbots: Ideal for building conversational agents that need to operate effectively in both English and Japanese.
• Instruction-Based Task Execution: Suitable for applications requiring the model to perform tasks based on natural language instructions, such as translation, summarization, or question answering.
• Research and Development: Provides a strong foundation for further fine-tuning or research into bilingual large language models, particularly those focused on Japanese language processing.