We are excited to introduce Hrida-T2SQL-3B-V0.1, our first open-source Text-to-SQL model fine-tuned on the Phi 3 architecture. The model converts natural language questions into precise SQL queries with exceptional accuracy and efficiency, revolutionizing database interaction. It enhances data analysis and exploration, making them more intuitive and effective. Additionally, Hrida-T2SQL-3B-V0.1 can be seamlessly integrated into production environments to automate database queries and streamline data processing workflows.
Why Hrida-T2SQL-3B-V0.1?
In today's data-driven world, the ability to query databases efficiently is crucial. However, not everyone is proficient in SQL. Hrida-T2SQL-3B-V0.1 bridges this gap by allowing users to ask questions in natural language and receive accurate SQL queries in response. This innovation opens up new possibilities for data analysis and accessibility, making it easier for anyone to leverage the power of databases.
To ensure Hrida-T2SQL-3B-V0.1 excels in real-world applications, we conducted an extensive benchmarking process. We carefully curated 50 questions that are outside training data. These questions encompass a diverse range of SQL queries and scenarios, facilitating a comprehensive evaluation of the model's performance across various query types and complexities. Each question was meticulously selected to represent a broad spectrum of SQL query structures and database interaction scenarios, thereby presenting unique challenges to the model.
To ensure fairness and objectivity, we had another LLM, Mistral 7B to judge the outputs and rate them on a scale of 1 to 10. Here’s how Hrida-T2SQL-3B-V0.1 stacks up against other leading models:
- Hrida-T2SQL - 3B: 9.26
- CodeQwen - 7B: 8.98
- Stable Code - 3B: 8.28
- DuckDB-NSQL - 7B: 6.34
- SQLCoder - 7B: 5.54
Our model, Hrida-T2SQL-3B-V0.1, exhibits superior performance with an impressive average score of 9.26. This underscores its accuracy and reliability, surpassing many existing solutions in the market. With its precision in translating natural language queries into SQL, our model empowers users to interact with databases more intuitively and efficiently.
Conclusion
Hrida-T2SQL-3B-V0.1 represents a leap forward in making database interactions more user-friendly and accessible. Whether you are a data scientist, business analyst, or someone who regularly works with databases, our model will make your life easier by eliminating the need for deep SQL knowledge.
We invite you to explore Hrida-T2SQL-3B-V0.1 and experience the future of data interaction. Stay tuned for more updates and enhancements as we continue to innovate.
For more information and to access the model, visit our Huggingface page.