SakanaAI/Llama-3-Karamaru-v1
Llama-3-Karamaru-v1 is a conversational AI model developed by Sakana AI, fine-tuned from Llama-3-ELYZA-JP-8B. It specializes in responding in the style of Edo-period Japanese, having undergone continual pretraining on a custom dataset of 25 million characters from that era. This model allows users to engage with historical Japanese language and cultural context, serving as a tool for research, education, and cultural exploration.
Loading preview...
Overview
SakanaAI's Llama-3-Karamaru-v1 is a unique conversational AI model designed to interact in the classical Japanese style of the Edo period. Developed by Sakana AI and fine-tuned from Llama-3-ELYZA-JP-8B, Karamaru bridges modern Japanese queries with historical linguistic responses.
Key Capabilities
- Edo-period Japanese Style: Responds to modern Japanese input with output reflecting the worldview and cultural context of the Edo era.
- Specialized Training Data: Continually pretrained on a custom dataset of approximately 25 million characters from the Edo period, including human-transcribed text and AI-based kuzushiji OCR from historical books.
- Cultural Exploration: Offers a distinct way to explore and engage with Japan's historical language and thought.
Intended Use
Karamaru is primarily intended for research, education, and cultural exploration, providing a tool to connect the past with the present through language. Users can experience a unique linguistic and cultural immersion.
Limitations
Due to its training on historical Edo-period texts, the model may reflect the social norms, values, and biases of that time. This could lead to responses that are considered inappropriate, outdated, or offensive by modern standards, requiring users to exercise mindfulness, especially in public-facing applications.
Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.