This project is based on the InternLM2.5 series model and aims to create a generative dialogue system with a dark, rebellious style. By combining Retrieval-Augmented Generation (RAG) technology, have built a model capable of generating dialogue filled with villainous traits, such as threats, provocation, and coldness. The system retrieves relevant text information from a pre-constructed knowledge base, enriching its outputs with variety and depth.
- Dark Style Dialogue Generation: The model generates responses with a dark, villainous tone, suitable for applications in games, movies, novels, etc.
- RAG-Enhanced Generation: By incorporating external knowledge sources, the model's responses are more diverse and insightful.
- Customizable Dialogue Style: This model is particularly well-suited for users interested in villainous characters, dark philosophy, historical figures, or similar themes.
First, clone the project to your local machine:
git clone https://github.com/yourusername/DarkIntern.git
Place the villain quotes, dark novel excerpts, and other relevant data in the data/ folder. Ensure the format of the data is suitable for model training. If you don't have an existing dark dataset, you can gather data from public resources or write your own.
If you need to fine-tune the model, you can use the train.py script. Run the following command to begin training:
python scripts/train.py --datadir data/ --outputdir model/ --batch_size 16 --epochs 3
This command will fine-tune the model using the training data in the data/ directory and save the trained model in the model/ folder.
After training the model, you can generate dialogue with a dark, villainous style using the generate.py script. You can provide a prompt or situation, and the model will generate a response with a villainous tone.
python scripts/generate.py --model_dir model/ --input "What do you think the ultimate fate of humanity will be?"
When generating dialogue, the model retrieves relevant villainous-style text based on the input. You can use the retrieve.py script to query the retrieval database:
python scripts/retrieve.py --query "How do villains view heroes?" --indexdir data/retrievaldata/
python scripts/generate.py --model_dir model/ --input "If you were the ruler of the world, what would you do?"
Output: “The world? It’s nothing but floating dust. I will make all bow before me in fear; their fear is the proof of my existence.”
python scripts/generate.py --model_dir model/ --input "How does one become a villain?"
Output: “First, you must cast aside all moral constraints. Then, you must learn to manipulate others' fears, becoming their darkest nightmare.”