core-3/kuno-royale-v2-7b

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Mar 1, 2024License:cc-by-nc-2.0Architecture:Transformer0.0K Open Weights Cold

kuno-royale-v2-7b is a 7 billion parameter merged language model developed by core-3, building upon SanjiWatsuki/Kunoichi-DPO-v2-7B and eren23/ogno-monarch-jaskier-merge-7b-OH-PREF-DPO. This model is specifically designed to strengthen roleplaying prose, offering improved performance in personal roleplay tests compared to its base model. It maintains a 4096-token context length and shows improved leaderboard metrics in specific categories like GSM8K and MMLU.

Loading preview...

Model Overview

kuno-royale-v2-7b is a 7 billion parameter merged language model created by core-3, aiming to enhance the roleplaying prose capabilities of its predecessor, SanjiWatsuki/Kunoichi-DPO-v2-7B. It integrates components from eren23/ogno-monarch-jaskier-merge-7b-OH-PREF-DPO, a high-scoring 7B model on the Open LLM Leaderboard. The merge was performed using LazyMergekit, combining layers from both base models.

Key Capabilities

  • Enhanced Roleplaying Prose: Specifically developed to improve the quality and depth of roleplaying interactions.
  • Improved Benchmarks: Demonstrates better performance in certain leaderboard metrics compared to SanjiWatsuki/Kunoichi-DPO-v2-7B, particularly in GSM8K and MMLU.
  • Silly Tavern Compatibility: Works effectively with the Noromaid template recommended for Kunoichi-7B, providing context and instruct configurations.

When to Use This Model

  • Roleplaying Applications: Ideal for scenarios requiring nuanced and engaging character interactions.
  • Creative Writing: Suitable for generating descriptive and immersive narrative content.
  • Personal Use Cases: Promising for individual roleplay tests and applications where prose quality is paramount.