Aratako/Qwen3-8B-ERP-v0.1
Aratako/Qwen3-8B-ERP-v0.1 is an 8 billion parameter language model developed by Aratako, fine-tuned specifically for role-playing scenarios. Based on Aratako/Qwen3-8B-NSFW-JP, this model excels at generating character-driven dialogues and narratives. It is optimized for interactive storytelling and immersive role-play applications, supporting a 32768 token context length.
Loading preview...
Overview
Aratako/Qwen3-8B-ERP-v0.1 is an 8 billion parameter model developed by Aratako, specifically fine-tuned for enhanced role-playing capabilities. It is built upon the base model Aratako/Qwen3-8B-NSFW-JP.
Key Capabilities
- Dedicated Role-Playing: Optimized for generating character-specific responses and maintaining narrative consistency in role-play scenarios.
- System Prompt Customization: Users can define detailed character settings, dialogue situations, and world-building elements via the system prompt to guide the model's output.
- Chat Template Support: Utilizes a specific chat template (
<|im_start|>system,<|im_start|>user,<|im_start|>assistant) for structured conversational input, compatible withtokenizer.apply_chat_template. - Flexible Deployment: Examples provided for inference using both
ollama(GGUF version) andtransformerslibrary, demonstrating ease of integration.
Training Details
The model was trained with a learning rate of 1e-5, a cosine LR scheduler, and a global batch size of 128. It supports a maximum sequence length of 8192 and uses paged_adamw_8bit as the optimizer.
Good For
- Interactive Storytelling: Creating dynamic and engaging narratives where the model embodies a specific character.
- Character Simulation: Developing virtual characters that adhere to predefined personalities and dialogue styles.
- Creative Writing Assistance: Generating dialogue and plot points for role-playing games, interactive fiction, or character-driven content.
License
This model is released under the MIT License.