Hugging Face Overview
Overview
Sao10K/Fimbulvetr-11B-v2 is a 10.7 billion parameter language model developed by Sao10K, described as a "Solar-Based Model." This iteration, version 2, has undergone heavy testing and received positive feedback, indicating its readiness for general use. The model supports common prompting formats, making it adaptable for various applications.
Key Capabilities
- Flexible Prompting: Compatible with both Alpaca and Vicuna instruction formats, allowing for broad integration into existing workflows.
- General Text Generation: Designed for a wide range of text-based tasks, leveraging its 10.7 billion parameters for coherent and relevant outputs.
Recommended Usage
- Prompt Formats: Users can choose between Alpaca or Vicuna formats based on their preference or existing system requirements.
- Alpaca Format:
### Instruction:\n<Prompt>\n### Input:\n<Insert Context Here>\n### Response: - Vicuna Format:
System: <Prompt>\n\nUser: <Input>\n\nAssistant:
- Alpaca Format:
- SillyTavern Presets: "Universal Light" is recommended for optimal performance within the SillyTavern environment.
Development Status
The developer, Sao10K, has indicated that work on a v3 is currently halted, with focus shifting to dataset refinement. This suggests a commitment to improving the underlying data quality for future iterations.