🔄 Hybrid Synthesis: LLM + GAN
The headline feature of this release is hybrid synthesis, enabling you to generate both textual and tabular data in a unified, streamlined process:
- **LLM for Text Generation**: Use advanced language models to generate contextually rich and realistic text.
- **GAN for Tabular Data**: Produce high-quality numerical and categorical tabular data with GANs.
- **Seamless Integration**: Effortlessly combine both approaches into a single synthesis pipeline for consistent dataset generation.
🛠️ New `TextTabularSynth` Class
We’re introducing the `TextTabularSynth` class to manage this hybrid synthesis process. Key features include:
- Integration of both LLM and GAN setups.
- A unified interface to streamline synthetic data generation.
- Ensures consistency between textual and tabular data.
📊 Enhanced Customization
- **Diversity Control**: Adjust text diversity with the `diversity_threshold` parameter.
- **Flexible LLM Integration**: Select your preferred LLM models for both generation and judging.
- **GAN Configuration**: Detailed control over GAN architecture, training parameters, and more.
🤝 Feedback and Support
We value your input! If you encounter any issues or have suggestions for future enhancements, please feel free to open an issue on our [GitHub repository](https://github.com/IndoxGen/IndoxGen) or contact our support team.
Thank you for your continued support and trust in IndoxGen. We’re excited to see how the new hybrid synthesis capability will enhance your synthetic data generation projects!
Happy synthesizing!
The IndoxGen Team