This project offers a fine-tuned Qwen2.5-1.5B-Instruct model specialized for sensitive and adult-themed content, particularly excelling with Chinese text. It targets researchers and testers exploring AI's capabilities in these domains, providing a specialized tool for nuanced content generation.

How It Works

The model is derived from Qwen2.5-1.5B-Instruct, undergoing fine-tuning on a curated dataset rich in adult literature and sensitive topics, including moral, legal, erotic, and violent themes. Its primary advantage lies in its enhanced performance with Chinese text, stemming from the dataset's predominantly Chinese composition.

Quick Start & Requirements

• Install: pip install -r requirements.txt
• Run: python run_webUI.py to launch a graphical interface.
• Prerequisites: Python environment and dependencies specified in requirements.txt.
• Configuration: Recommended settings include default or higher TOP_P/TOP_K values, low or default temperature, and an empty system prompt.
• Dataset Links: Direct links to the specific datasets (Bad Data, Toxic-All, Erotic Literature Collection) are mentioned but not provided in the excerpt.

Highlighted Details

• Specialized fine-tuning on adult literature and sensitive datasets.
• Optimized performance for Chinese text due to dataset composition.
• Open-sourced on Hugging Face for research and testing purposes.

Maintenance & Community

Information regarding notable contributors, sponsorships, partnerships, or community channels (e.g., Discord, Slack) is not present in the provided README excerpt.

Licensing & Compatibility

A formal license is not specified. The project is designated strictly for "research and testing use only," with users bearing full responsibility for legal compliance, implying potential restrictions for commercial applications.

Limitations & Caveats

This model is strictly designated for research and testing. Users must ensure compliance with all applicable local laws and regulations, assuming full responsibility for their use. The underlying dataset contains potentially sensitive, controversial, violent, erotic, illegal, and immoral content, necessitating deployment within controlled environments.