Shumatsurontek/Qwen3.5-4B-neo
Shumatsurontek/Qwen3.5-4B-neo is a 4.5 billion parameter language model fine-tuned by Shumatsurontek from unsloth/Qwen3.5-4B. Optimized for text-to-SQL generation, it excels at converting natural language questions and database schemas into SQL queries. This model is specifically designed for analytical and read-only SQL tasks, offering a specialized solution for database interaction.
Loading preview...
Qwen3.5-4B-neo: Text-to-SQL Fine-tuned Model
This model, developed by Shumatsurontek, is a specialized version of the unsloth/Qwen3.5-4B base model, fine-tuned for text-to-SQL generation. It leverages Supervised Fine-Tuning (SFT) with LoRA adapters, which are merged into the base weights for efficient deployment. The training utilized the Shumatsurontek/neo-sql-reasoning-combined dataset, focusing on converting natural language questions and database schemas into accurate SQL queries.
Key Capabilities
- Text-to-SQL Generation: Translates natural language questions and provided database schemas into SQL queries.
- Specialized for Analytical Queries: Best suited for generating read-only SQL queries for data analysis.
- Efficient Deployment: LoRA adapters are merged into the base model, simplifying integration.
- Apache 2.0 Licensed: Provides flexibility for commercial and open-source projects.
Good For
- Database Interaction: Automating the generation of SQL queries from user input.
- Data Analysis Tools: Integrating natural language interfaces for querying databases.
- Educational Purposes: Demonstrating text-to-SQL capabilities and fine-tuning techniques.
Limitations
- Out of Scope: Not intended for DDL/DML generation (CREATE, DROP, INSERT, UPDATE, DELETE), multi-database queries, or production use without human review of generated SQL.
- General Reasoning Drop: Benchmarks show a slight decrease in general reasoning tasks (MMLU, MMLU: STEM, MMLU: HUMANITIES, MMLU: SOCIAL SCIENCES) compared to the baseline, indicating its specialized nature.