Novaciano/ProtoSex-RP-Gemma2-2B
Novaciano/ProtoSex-RP-Gemma2-2B is a 2.6 billion parameter Gemma2-based language model, fine-tuned for immersive roleplay with an emphasis on character consistency, dialogue flow, and expressive narration. Developed by Novaciano, this model is designed to provide detailed and engaging roleplaying experiences, including optional adult depth. It leverages a merged architecture to enhance sensory detail and reduce refusals, making it suitable for interactive narrative generation.
Loading preview...
Overview
Novaciano/ProtoSex-RP-Gemma2-2B is a 2.6 billion parameter language model built on the Gemma2 architecture, specifically designed for immersive roleplay scenarios. This prototype model focuses on delivering detailed and engaging narrative experiences, including the capacity for adult-themed content. It was created using the Model Stock merge method, combining several specialized Gemma2-based models to achieve its unique capabilities.
Key Capabilities
- Immersive Roleplay: Optimized for deep, character-consistent, and expressive narrative generation.
- Dialogue Flow: Enhances natural and coherent conversational exchanges within roleplay.
- Sensory Detail & Intimacy: Incorporates rich descriptive elements and the ability to convey intimate scenarios.
- Reduced Refusals: Engineered to minimize uncooperative responses while maintaining stable instruction-following.
- Gemma2 Base: Benefits from the underlying architecture of Gemma2, providing a strong foundation for language understanding and generation.
Merge Details
This model is a merge of three distinct Gemma2-based models, with ChiKoi7/Gemma-2-Llama-Swallow-2b-it-v0.1-Heretic serving as the base. The merged components include:
- ArliAI/Gemma-2-2B-ArliAI-RPMax-v1.1: Contributes to primary roleplay driving, character consistency, and dialogue flow.
- akashgoel-id/En_RP_DPO-gemma2_2b_64X32_test: Improves roleplay alignment through DPO, reducing refusals without compromising narrative quality.
- TheDrummer/Gemmasutra-Mini-2B-v1: Adds sensory detail, intimacy, and expressive narration to the model's output.
Recommended Inference Parameters
For optimal performance, the following inference parameters are suggested:
- Temperature: 0.6 – 0.75
- Top-p: 0.9
- Top-k: 50 – 80
- Repetition penalty: 1.1
- Presence penalty: 0.2 – 0.4