T5-small-awesome-text-to-sql — ONNX

ONNX export of t5-small-awesome-text-to-sql (60M parameters) with encoder-decoder architecture and KV cache support.

A lightweight T5-small model fine-tuned for text-to-SQL generation. Accepts database schemas as CREATE TABLE DDL statements and generates SQL queries from natural language questions.

Converted for use with inference4j, an inference-only AI library for Java.

Original Source

• Repository: cssupport/t5-small-awesome-text-to-sql
• License: Apache 2.0

Usage with inference4j

try (var sqlGen = T5SqlGenerator.t5SmallAwesome().build()) {
    String sql = sqlGen.generateSql(
        "How many employees are in each department?",
        "CREATE TABLE employees (id INT, name VARCHAR, department VARCHAR, salary INT); "
        + "CREATE TABLE departments (id INT, name VARCHAR)");
    System.out.println(sql);
}

Schema Format

The model expects standard SQL DDL as schema input:

CREATE TABLE employees (id INT, name VARCHAR, salary INT); CREATE TABLE departments (id INT, name VARCHAR)

For higher accuracy on complex multi-table queries with JOINs, GROUP BY, and subqueries, consider the larger
T5-LM-Large-text2sql-spider (0.8B parameters).

Model Details

License

This model is licensed under the Apache License 2.0. Original model by cssupport.