cycloneboy/SLM-SQL-0.5B

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:0.5BQuant:BF16Ctx Length:32kPublished:Jul 31, 2025License:cc-by-nc-4.0Architecture:Transformer0.0K Open Weights Warm

SLM-SQL-0.5B by cycloneboy is a 0.5 billion parameter small language model (SLM) specifically fine-tuned for Text-to-SQL tasks, building upon the Qwen2.5-Coder-0.5B-Instruct base model. It leverages supervised fine-tuning (SFT) and reinforcement learning-based post-training (GRPO) with a corrective self-consistency approach to enhance its logical reasoning for SQL generation. This model is designed to overcome the limitations of SLMs in Text-to-SQL, offering advantages in inference speed and suitability for edge deployment, achieving 56.87% execution accuracy on the BIRD development set.

Loading preview...

SLM-SQL: Small Language Models for Text-to-SQL

SLM-SQL-0.5B is a 0.5 billion parameter model developed by cycloneboy, specifically optimized for translating natural language questions into SQL queries (Text-to-SQL). Unlike larger LLMs, this model focuses on the advantages of Small Language Models (SLMs) such as faster inference and suitability for edge deployment, while significantly improving their Text-to-SQL performance.

Key Capabilities

  • Enhanced Text-to-SQL Performance: Achieves 56.87% execution accuracy (EX) on the BIRD development set, a notable improvement for its size class.
  • Advanced Training Techniques: Utilizes supervised fine-tuning (SFT) and reinforcement learning-based post-training (GRPO) on custom datasets like SynSQL-Think-916K and SynSQL-Merge-Think-310K.
  • Corrective Self-Consistency: Employs an inference strategy that further refines SQL query generation.
  • Efficient Deployment: Designed for scenarios where computational resources are limited, offering a balance between performance and efficiency.

Good for

  • Edge Computing: Ideal for deploying Text-to-SQL capabilities on devices with constrained resources.
  • Applications Requiring Fast Inference: Suitable for real-time or near real-time natural language to SQL conversion.
  • Database Interaction: Enables users to query databases using natural language, even with a compact model footprint.
  • Research in SLM Optimization: Demonstrates effective post-training techniques to boost SLM performance in complex logical reasoning tasks.

Finetunes
1 models