gravelight-studio/EstopianMaid-13B
EstopianMaid-13B by gravelight-studio is a 13 billion parameter language model built by merging several LLaMA2-based models, including TimeCrystal-l2-13B and Thespis-13b-DPO-v0.7. This model is specifically designed for character-driven interactions, excelling at maintaining character consistency and coherency in multi-character scenarios. With a 4096-token context length, it is optimized for creating new scenarios and adhering to character cards, making it suitable for roleplay and interactive storytelling applications.
Loading preview...
EstopianMaid-13B: Character-Focused Language Model
EstopianMaid-13B is a 13 billion parameter language model developed by gravelight-studio, designed with a strong emphasis on character interaction and narrative consistency. This model is a merge of several specialized 13B models, including BlueNipples/TimeCrystal-l2-13B, cgato/Thespis-13b-DPO-v0.7, KoboldAI/LLaMA2-13B-Estopia, NeverSleep/Noromaid-13B-0.4-DPO, and Doctor-Shotgun/cat-v1.0-13b, leveraging their combined strengths.
Key Capabilities
- Character Adherence: Excels at consistently sticking to defined character cards and personas.
- Multi-Character Coherency: Maintains narrative and character coherency even in complex settings involving multiple characters.
- Scenario Generation: Capable of creating and developing new scenarios effectively.
- Thespis Feature Integration: Incorporates features from Thespis-13b-DPO-v0.7, enhancing its dialogue and roleplay capabilities.
Recommended Usage
This model is particularly well-suited for applications requiring robust character interaction and dynamic storytelling. It uses the Alpaca prompt template for instructions. Recommended settings for optimal performance include a temperature of 0.7, Min-P of 0.3, Top P of 1, and a repetition penalty of 1.10, with a generation length of 256 tokens.