slim-sql-onnx

slim-sql-onnx is a small specialized function calling model that takes as input a table schema and a natural language query, and outputs a SQL statement that corresponds to the query, and can be run against a database table. This is a very small text-to-sql model designed for reasonable accuracy on single tables and relatively straightforward queries, and for easy integration into multi-step processes.

This is an ONNX int4 quantized version of slim-sql-1b-v0, providing a very fast, very small inference implementation, optimized for AI PCs using Intel GPU, CPU and NPU.

Model Description

• Developed by: llmware
• Model type: tinyllama
• Parameters: 1.1 billion
• Model Parent: llmware/slim-sql-1b-v0
• Language(s) (NLP): English
• License: Apache 2.0
• Uses: Text-to-SQL conversion
• RAG Benchmark Accuracy Score: NA
• Quantization: int4

Model Card Contact

llmware on github

llmware on hf

llmware website