Gemini AI Studio Fine-Tuning Dataset Creator

Built for Gemma models fine tuning

A PyQt5-based GUI application designed to generate structured fine-tuning datasets in the Dolly format using Google Gemini AI. This tool is particularly useful for creating high-quality training data from technical manuals or any subject-specific content. Features

Dolly Format Dataset Generation: Creates training data following the Dolly instruction format, which includes instruction, context, response, and category fields Bulk Context & Category Labeling: Add consistent context and category labels across all entries for efficient dataset organization Flexible Subject Support: While designed for technical manuals, it works with any subject matter - just set the "machine name" to your topic of interest (e.g., "green tree frogs", "quantum physics") Reference Q&A Integration: Input previous Q&A pairs to avoid duplicates and guide new question generation Dark Mode Interface: Material Design-inspired dark theme for comfortable extended use Multiple Gemini Models: Support for various Gemini AI models with real-time rate limit tracking File Management: Import source text from files and export structured datasets in JSON or text format

Rate Limiting & API Usage The application includes real-time tracking of API usage with built-in rate limiting for different Gemini models. This helps prevent hitting API limits while working with larger datasets. Installation & Setup

Clone the repository Install required packages

Set your GEMINI_API_KEY environment variable Run the application

Dataset Format The tool now generates data in the Dolly instruction format: