Azzedde/llama3.2-3b-sql-expert-1-epoch-gguf

Model Card for Llama3.2-3B-SQL-Expert-1Epoch (GGUF)

Model Details

• Model Name: Llama3.2-3B-SQL-Expert-1Epoch (GGUF)
• Developer: Azzedine (GitHub: Azzedde)
• Base Model: Meta-Llama-3.1-3B-Instruct
• Format: GGUF
• License: Apache 2.0
• Language: English
• Purpose: SQL query generation from natural language input

Model Sources

• Repository: Hugging Face

Use Cases

• Generating SQL queries from natural language prompts
• Assisting developers with database management
• Supporting SQL education and analytics automation

Limitations

• May generate suboptimal or incorrect SQL queries
• Does not validate database schema relationships
• Primarily supports English

How to Use

from llama_cpp import Llama  

model = Llama("llama3.2-3b-sql-expert-1-epoch.gguf")  

prompt = """### Schema:
CREATE TABLE employees (id INT PRIMARY KEY, name VARCHAR, salary DECIMAL, department_id INT);

### Question:
Find the average salary per department.

### SQL Query:
"""

output = model(prompt, max_tokens=64)  
print(output)

Training Details

• Data: Public SQL datasets + synthetic SQL queries
• Epochs: 1
• Fine-tuned with: Unsloth
• Hardware: Tesla T4

Citation

@article{llama3.2-3B-SQL-Expert,
  author    = {Azzedde},
  title     = {Llama3.2-3B-SQL-Expert: An SQL Query Generation Model (GGUF)},
  year      = {2025},
  url       = {https://huggingface.co/Azzedde/llama3.2-3b-sql-expert-1-epoch}
}