ParasiticRogue/RP-Stew-v4.0-34B
ParasiticRogue/RP-Stew-v4.0-34B is a 34 billion parameter experimental merge model, based on the Capybara architecture, designed to enhance roleplay and creative writing capabilities. With a context length of 32768 tokens, this model integrates several specialized 34B models to improve its performance in generating nuanced and engaging narrative content. It is primarily intended for advanced roleplaying and story generation applications, offering a refined output compared to previous iterations.
Loading preview...
RP-Stew-v4.0-34B Overview
RP-Stew-v4.0-34B is an experimental 34 billion parameter merged language model developed by ParasiticRogue, building upon the Capybara base architecture. This iteration aims to significantly enhance the model's capabilities, particularly in the realm of roleplaying and creative content generation, showing promising results in initial tests compared to its predecessors.
Key Capabilities & Features
- Enhanced Roleplay: Specifically designed and merged from models known for their roleplaying strengths, aiming for more coherent and engaging narrative outputs.
- Extensive Context Window: Supports a context length of 32768 tokens, allowing for longer and more complex conversational or narrative threads.
- Model Merging Strategy: Integrates several high-performing 34B models, including Nous-Capybara-34B, Tess-2.0-Yi-34B-200K, bagel-dpo-34b-v0.5, PiVoT-SUS-RP, NyakuraV2-34B-Yi-Llama, NeverSleep/CausalLM-RP-34B, and Yi-34b-200K-AEZAKMI-RAW-TOXIC-2702, to combine their respective strengths.
- Vicuna-1.1 Chat Format: Utilizes the
Chat-Vicuna-1.1prompt format for structured interactions.
Recommended Usage
- Roleplaying Scenarios: Ideal for generating detailed and immersive roleplay responses.
- Creative Writing: Suitable for tasks requiring imaginative storytelling, character development, and narrative continuation.
- Experimental Applications: Given its experimental nature, it's well-suited for developers and enthusiasts looking to explore advanced language model capabilities in creative domains.
Users should note that trust-remote-code must be enabled due to the Capybara base model, a setting the developer plans to address in future updates if performance remains strong.