Model Overview

This model, developed by ShacharNar, is a specialized fine-tuned version of the Qwen2.5-Coder-3B architecture, featuring 3.1 billion parameters and a 32768-token context window. It has been specifically trained using the TRL framework to excel in SQL-related tasks, with a particular focus on T-SQL generation.

Key Capabilities

• T-SQL Generation: Optimized for generating Transact-SQL queries and statements.
• SQL Query Answering: Designed to provide answers to SQL-related questions, especially when responses are expected within defined delimiters.
• Code-focused Fine-tuning: Builds upon the Qwen2.5-Coder-3B base, enhancing its capabilities for code understanding and generation.

Training Details

The model underwent Supervised Fine-Tuning (SFT) using the TRL library, ensuring its specialized performance in SQL contexts. The training process utilized specific versions of TRL (0.19.0), Transformers (4.57.6), Pytorch (2.6.0), Datasets (3.5.0), and Tokenizers (0.22.0).

Good for

• Developers and data professionals working with T-SQL databases.
• Automating T-SQL query generation.
• Applications requiring precise, delimited answers to SQL-related questions.