timbossm/qwen2.5-3B-sql-mgpu-bi-ft

TEXT GENERATIONConcurrency Cost:1Model Size:3.1BQuant:BF16Ctx Length:32kPublished:May 10, 2025License:apache-2.0Architecture:Transformer Open Weights Cold

The timbossm/qwen2.5-3B-sql-mgpu-bi-ft model is a 3.1 billion parameter Qwen2.5-based language model, fine-tuned for Text-to-SQL generation. Developed by timbossm, it specializes in converting natural language instructions and database schema context into SQL queries. This model is an iteration of Ellbendls/Qwen-2.5-3b-Text_to_SQL, further optimized using LoRA on a custom dataset for enhanced SQL generation capabilities.

Loading preview...

Overview

This model, timbossm/qwen2.5-3B-sql-mgpu-bi-ft, is a specialized 3.1 billion parameter language model designed for Text-to-SQL generation. It is a further fine-tuned version of Ellbendls/Qwen-2.5-3b-Text_to_SQL, which itself is based on Qwen/Qwen2.5-3B-Instruct.

Key Capabilities

  • Text-to-SQL Conversion: Translates natural language prompts into executable SQL queries.
  • Schema-Aware Generation: Utilizes provided database schema (DDL scripts) as context to generate accurate SQL.
  • Fine-tuned Performance: Enhanced through LoRA (Parameter-Efficient Fine-Tuning) on a custom dataset, building upon its base models trained on gretelai/synthetic_text_to_sql and timbossm/sql_bi__b_db.

Intended Use Cases

  • SQL Query Generation: Ideal for creating SQL queries from natural language instructions.
  • Data Analysis Assistance: Supports users in interacting with databases for data retrieval and analysis.
  • Educational Purposes: Can be used as a tool for learning and practicing SQL.

Limitations

  • Context Dependency: Generation quality is highly dependent on the accuracy and completeness of the provided database schema context.
  • SQL Dialect Specificity: Performance on SQL dialects not represented in its training data is not guaranteed.
  • Scope: Not intended for general dialogue, code generation in other languages, or complex multi-step reasoning tasks. It inherits limitations and potential biases from its foundational Qwen models.