Hugging Face Overview
Hrida-T2SQL-3B-V0.1: Text-to-SQL SLM
Hrida-T2SQL-3B-V0.1 is a 4 billion parameter Small Language Model (SLM) developed by HridaAI, specifically fine-tuned for Text-to-SQL tasks. It is based on the Microsoft/Phi-3-mini-4k-instruct architecture, leveraging its compact size and efficiency for specialized applications.
Key Capabilities
- Natural Language to SQL Conversion: Translates user queries in natural language into executable SQL statements.
- Dialect Specification: Supports specifying various SQL dialects (e.g., MySQL, PostgreSQL, SQL Server) to ensure syntax conformity.
- Schema-Aware Generation: Utilizes provided database schemas (table names, column names, data types) to generate accurate and contextually relevant SQL queries.
- Instruction-Guided Output: Employs a structured prompt template including system instructions, dialect, context, and input to guide SQL generation.
Prompt Template Structure
The model uses a specific prompt format to optimize performance, including:
### Instruction: System prompt to guide the model.### Dialect: Optional SQL variant specification.### Context: Database schema details.### Input: User's natural language query.### Response: Expected SQL query output.
Good For
- Automating database query generation from user input.
- Integrating natural language interfaces with SQL databases.
- Applications requiring efficient and accurate SQL code generation from text.