zarakiquemparte/kuchiki-l2-7b
The zarakiquemparte/kuchiki-l2-7b is a 7 billion parameter merged language model, combining Nous Hermes Llama2 7b, Airoboros L2 7B GPT4 2.0, and LimaRP LLama2 7B Lora. This model leverages the strengths of its base components, offering a 4096-token context length. It is designed for diverse conversational and instruction-following tasks, supporting both Alpaca 2 and LimaRP instruction formats.
Loading preview...
Model Overview
Kuchiki L2 7b is a 7 billion parameter language model created by zarakiquemparte through a strategic merge of several established models. It primarily uses Nous Hermes Llama2 7b (70%) and Airoboros L2 7B GPT4 2.0 (30%) as its foundational base, further enhanced by a merge with LimaRP LLama2 7B Lora.
Key Characteristics
- Merged Architecture: Combines the capabilities of Nous Hermes, Airoboros, and LimaRP for a versatile performance profile.
- Instruction Format Compatibility: Supports both the Alpaca 2 and LimaRP instruction formats, making it adaptable to various prompting styles.
- Reproducible Merging Process: The merging process was conducted using custom scripts, which are publicly available, allowing for reproducibility and transparency.
Usage Considerations
This model is suitable for general instruction-following and conversational applications, benefiting from the diverse training data of its constituent models. Users should be aware that, like many merged models, it is not intended for providing factual information or advice. Quantized versions (GGUF and GPTQ) are available from TheBloke for optimized deployment.