sophosympatheia/Sophos-eva-euryale-v1.0
Sophos-eva-euryale-v1.0 by sophosympatheia is a 70 billion parameter merged language model, combining EVA-LLaMA-3.33-70B-v0.0 and L3.3-70B-Euryale-v2.3 using the SLERP method. This model is specifically designed and optimized for roleplaying and storytelling tasks, excelling in generating creative and consistent narrative content. It is an uncensored model, providing flexibility for various creative applications.
Loading preview...
Model Overview
Sophos-eva-euryale-v1.0 is a 70 billion parameter language model developed by sophosympatheia. It is a merge of two distinct models: EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.0 and Sao10K/L3.3-70B-Euryale-v2.3, created using the SLERP merge method. This model is explicitly designed to be uncensored, offering broad creative freedom.
Key Capabilities
- Roleplaying and Storytelling: The primary focus and strength of this model lie in generating engaging and consistent content for roleplaying and storytelling scenarios.
- Uncensored Output: Provides uncensored responses, allowing for a wider range of narrative themes and content.
Usage Recommendations
The developer provides specific sampler tips for optimal performance, particularly for creative generation:
- Min-P: Recommended settings between 0.02 and 0.1.
- DRY Repetition Penalty: Suggested multiplier around 0.5 with a base between 1.5 and 2.0.
- Temperature: Experiment with settings in the 0.8 - 1.2 range to balance creativity and coherence.
Licensing and Disclaimer
This model incorporates weights from multiple LLMs, leading to uncertain licensing terms. Users are advised to acknowledge potential legal risks, especially for commercial applications, and consult legal counsel for compliance. The Llama 3.3 Community License Agreement applies to one of its base components.