README.md

---
title: SQL Generation
emoji: 🦀
colorFrom: red
colorTo: gray
sdk: gradio
sdk_version: 5.9.1
app_file: app.py
pinned: false
license: mit
---

# SQL Generation 🦀

Welcome to the **SQL Generation** Gradio application! This tool leverages advanced machine learning models to assist in generating SQL queries based on natural language inputs. Whether you're a developer, data analyst, or just curious about SQL, this app aims to simplify the process of crafting SQL queries.

## Features

- **Natural Language to SQL**: Convert plain English descriptions into SQL queries.
- **Multiple Datasets**: Trained on diverse datasets to handle various SQL generation tasks.
- **User-Friendly Interface**: Built with Gradio for an intuitive and interactive experience.

## Installation

To run this application locally, ensure you have Python 3.10 or higher installed. Then, install the required dependencies:

```bash
pip install gradio transformers datasets
```
## Usage
After installing the dependencies, you can start the application by running:

```bash
python app.py
```
This will launch a local server. Open your browser and navigate to http://127.0.0.1:7860 to access the interface.

### Datasets Used
The model has been trained on the following datasets:
- b-mc2/sql-create-context: Provides context for SQL query generation.
- TuneIt/o1-python: Offers examples of Python code snippets.
- HuggingFaceFW/fineweb-2: Includes various language models for fine-tuning.
- sentence-transformers/embedding-training-data: Supplies data for training sentence embeddings.

## Model
The application utilizes the distilbert-base-uncased model from Hugging Face, known for its efficiency and performance in natural language processing tasks.

## License
This project is licensed under the MIT License.

## Acknowledgments
- **Gradio** for providing an easy-to-use interface for machine learning models.
- **Hugging Face** for hosting the pre-trained models and datasets.
- **Datasets** for offering a wide range of datasets for training and evaluation.

For more information, refer to the **Gradio** documentation.