Hugging Face Overview
Model Overview
L3.3-GeneticLemonade-Final-70B is a 70 billion parameter model developed by zerofata, specifically engineered for roleplay (RP) and creative writing. It is the second iteration in the Genetic Lemonade series, built upon the Llama-3.3-70B-Instruct base model using the SCE merge method.
Key Characteristics
- Uncensored Output: This "Final" version is explicitly designed to be fully uncensored, offering complete freedom for creative expression in RP and writing scenarios.
- Creative Writing Focus: The model aims to strike a balance between writing style, creativity, and intelligence, making it suitable for generating engaging narratives and dialogues.
- Merge Strategy: The model's base was constructed from high-performing models trained on diverse datasets to foster natural creativity and reduce data overlap. A second merge step then imparted specific RP and creative writing knowledge.
- Context Length: It supports a context length of 32768 tokens.
When to Use This Model
- Unrestricted Roleplay: Ideal for users requiring a model that provides fully uncensored responses for roleplaying scenarios.
- Creative Writing: Suitable for generating creative content where a balance of writing style and imaginative output is desired.
- SillyTavern Integration: Recommended for use with SillyTavern, with specific sampler settings provided (Temp: 0.9-1.0, MinP: 0.03-0.05, Dry: 0.8, 1.75, 4) and Llama-3-Instruct-Names for instruct prompting (with "System same as user" unchecked).