Bilingual-SQL-Coder

GitHub Repo | Paper/Docs

A comprehensive Text-to-SQL framework supporting bilingual (English & Chinese) tasks.

---

English | 中文

English

Model Description

Bilingual-SQL-Coder is a fine-tuned model specialized in Text-to-SQL tasks. It supports both English and Chinese queries and is capable of generating complex SQL queries for various databases.

This model is part of the Bilingual-SQL-Coder Project, which provides an end-to-end solution for data processing, training, and evaluation.

• Task: Text-to-SQL (Natural Language to SQL)
• Language: English & Chinese (Bilingual)
• Training Method: SFT (Supervised Fine-Tuning) with DoRA
• Base Model: Qwen3-4b-Instruct

Dataset & Training

• Datasets: Trained on a combination of Spider (English), CSpider (Chinese), and synthetic bilingual datasets.
• Pipeline: Data processed using the data_expert module from the main repository, including syntax fixing and augmentation.

Citation

If you use this model, please cite:

@software{bilingual_sql_coder,
  title={Bilingual-SQL-Coder: A Text-to-SQL Framework},
  author={Tianyu Gao, Deyang Wang, Yiwen Ma},
  year={2025},
  url={[https://github.com/GodRayyy/Bilingual-SQL-Coder](https://github.com/GodRayyy/Bilingual-SQL-Coder)}
}

---

中文

模型介绍

Bilingual-SQL-Coder 是一个专注于 Text-to-SQL（文本转SQL）任务的微调模型。它不仅支持英文，还对中文查询进行了深度优化，能够根据数据库 Schema 生成准确的 SQL 语句。

该模型是 Bilingual-SQL-Coder 开源项目 的一部分。

• 任务: Text-to-SQL (自然语言转 SQL)
• 语言: 中文 & 英文 (双语支持)
• 训练方法: SFT (有监督微调) 结合 DoRA 技术
• 基座模型: Qwen3-4b-Instruct

数据与训练

• 数据来源: 训练数据包含 Spider (英文)、CSpider (中文) 以及通过项目 pipeline 合成的双语数据。
• 数据处理: 使用了项目中的 data_expert 模块进行语法修正、清洗和增强。

引用

如果您在研究中使用了本模型，请引用：

@software{bilingual_sql_coder,
  title={Bilingual-SQL-Coder: A Text-to-SQL Framework},
  author={Tianyu Gao, Deyang Wang, Yiwen Ma},
  year={2025},
  url={[https://github.com/GodRayyy/Bilingual-SQL-Coder](https://github.com/GodRayyy/Bilingual-SQL-Coder)}
}