Vezora/Narwhal-7b
Vezora/Narwhal-7b is a 7 billion parameter model created by Vezora, built from a blend of Stable Beluga, MegaCoder, Wizard-Math, and Llama Chat7b. This model demonstrates remarkable performance in mathematical tasks and maintains robust query response capabilities. Optimized with Llama v2 prompting, it is noted for its strong performance within its category. Due to underlying training data, this model is explicitly not intended for commercial use.
Loading preview...
Narwhal-7b Overview
Narwhal-7b is a 7 billion parameter model developed by Vezora, engineered through a unique blend of several foundational models. It incorporates 60% Stable Beluga and 40% MegaCoder, further enhanced with 40% Wizard-Math and 40% Llama Chat7b. This specific combination aims to leverage the strengths of each component.
Key Capabilities
- Mathematical Proficiency: The model exhibits strong performance in mathematical tasks, a direct result of its specialized blend including Wizard-Math.
- Robust Query Response: It maintains a solid ability to respond to a wide range of queries, benefiting from its diverse training components.
- Optimized Prompting: Testing revealed that Llama v2 prompting yields superior results, indicating an optimization in its response generation, likely due to its integration as the final merge step.
Important Considerations
- Non-Commercial Use: Due to the inclusion of datasets originating from OpenAI in its training, Narwhal-7b is explicitly restricted from commercial use. Users must be aware of the commercial licenses associated with its underlying models.
- Performance: The model is positioned as one of the best-performing in its category, particularly for its mathematical and general query capabilities.
- Benchmarks: Comprehensive benchmarking details are anticipated to be released soon, which will provide further insights into its performance metrics.