Description
QwerySmith 1.0 is a Text-to-SQL model developed to facilitate the conversion of natural language queries into SQL commands. Built on the unsloth/Qwen3-4B model, it has been fine-tuned using QLoRA with Unsloth, leveraging 10,000 text-to-SQL context pairs. This model is particularly useful for database administrators and developers who need to translate user-friendly language into precise SQL queries.
Read More
The model demonstrates impressive benchmark results, with a valid SQL rate of up to 99.5% in few-shot scenarios and an exact match rate of 87.5% when fine-tuned. Its execution accuracy ranges from 34.6% to 88.5% depending on the dataset and system configuration. These metrics indicate its reliability and efficiency in producing correct SQL commands.
QwerySmith 1.0 is designed to run queries in SQLite without errors, ensuring that the predicted and gold queries return the same rows. It is particularly beneficial for environments where SQL dialects vary, although it is optimized for SQLite.
The model is open source, aligning with Hugging Face's mission to democratize artificial intelligence through open science. It is suitable for industries reliant on data management and analysis, providing a robust tool for transforming textual data into actionable SQL queries.
- CategoryAI Models & LLMs
- Pricing ModelOpen source
- DeploymentCloud
- PlatformsWeb
QwerySmith 1.0 Highlights
Text-to-SQL conversion
Fine-tuning with QLoRA
High SQL validity rate
Exact match accuracy
Execution accuracy
Open source model
Based on Qwen3-4B
Optimized for SQLite
Getting Started with QwerySmith 1.0
Access model: Visit Hugging Face
Authenticate: Use your credentials
Set up environment: Prepare SQLite
Integrate via API: Connect model
Optimize: Fine-tune with QLoRA
QwerySmith 1.0's Use Cases
- Database Management
- Data Analysis
- SQL Learning
- Automated Reporting
- Business Intelligence
