XGenerationLab/XiYanSQL-QwenCoder-14B-2504
The XiYanSQL-QwenCoder-14B-2504 is a 14 billion parameter SQL generation model developed by XGenerationLab, optimized for converting natural language to SQL queries. It leverages a combination of fine-tuning and GRPO training strategies to achieve high efficiency and accuracy. This model supports multiple SQL dialects, including SQLite, PostgreSQL, and MySQL, and demonstrates strong generalization capabilities across various dialects and out-of-domain datasets, making it suitable for complex real-world SQL generation tasks.
Loading preview...
XiYanSQL-QwenCoder-14B-2504: Advanced SQL Generation Model
XiYanSQL-QwenCoder-14B-2504 is the latest 14 billion parameter iteration in XGenerationLab's series of SQL generation models, designed to convert natural language into SQL queries with enhanced performance and generalization. This version incorporates a novel training approach combining fine-tuning with GRPO (Generative Reinforcement Pre-training Optimization), which allows for efficient and accurate SQL generation without requiring a complex thinking process.
Key Capabilities
- Multi-Dialect Support: Generates SQL for mainstream dialects including SQLite, PostgreSQL, and MySQL.
- Improved Generalization: Excels on diverse dialects and out-of-domain datasets, validated through a real-world DW test set comprising thousands of complex queries.
- Optimized Performance: Achieves competitive results on benchmarks like BIRD and Spider, with notable improvements over previous versions and other models in its class, particularly on M-Schema and DDL formats.
- Efficient Training: Utilizes GRPO training for a balance of efficiency and accuracy in SQL generation.
Good for
- Text-to-SQL Applications: Ideal for systems requiring robust and accurate conversion of natural language questions into SQL queries.
- Database Interaction: Facilitating user interaction with databases across various SQL dialects.
- Complex Query Generation: Handling intricate and real-world SQL scenarios, as demonstrated by its performance on the DW test set.
- Developers and Data Analysts: Streamlining the process of writing SQL queries from natural language prompts.