sh0ck0r/Strawberrylemonade-L3-70B-v1.2-heretic: Decensored Roleplay LLM
This model is a 70 billion parameter, decensored version of sophosympatheia/Strawberrylemonade-L3-70B-v1.2, created using the Heretic v1.2.0 tool. It is a merge of multiple Llama 3-based models, including zerofata/L3.3-GeneticLemonade-Final-v2-70B, zerofata/L3.3-GeneticLemonade-Unleashed-v3-70B, deepcogito/cogito-v1-preview-llama-70B, and arcee-ai/Arcee-SuperNova-v1, utilizing the NuSLERP merge method. The primary goal of this model is to provide an uncensored experience, specifically tailored for roleplaying and storytelling with improved creativity and expressiveness.
Key Capabilities
- Decensored Output: Significantly reduced refusals (21/100 compared to 99/100 in the original model), allowing for unrestricted content generation.
- Enhanced Roleplaying & Storytelling: Designed to excel in immersive narrative development, character portrayal, and creative ideation.
- Configurable Output Style: Highly responsive to sampler settings like
min-p and temperature, enabling users to fine-tune for either reliable, coherent responses or more creative, unpredictable outputs. - Llama 3 Base: Benefits from the underlying architecture and capabilities of the Llama 3 family.
Good For
- Creative Writing Applications: Ideal for generating diverse and imaginative stories, scenarios, and dialogues.
- Roleplaying Bots and Simulations: Suited for developing characters with distinct personalities and engaging in dynamic, turn-based roleplay.
- Unrestricted Content Generation: For use cases where content moderation or refusal behavior is undesirable, particularly in creative or experimental contexts.
- Experimentation with Sampler Settings: Users who wish to explore the impact of various generation parameters on model output.