ChangeFormer Core Code

This repository contains the core code for ChangeFormer. The index_seleciton_evaluation/ChangeFormer directory includes the model definitions and training files for ChangeFormer and the ARQG directory contains the code for randomly generated queries. The index_seleciton_evaluation comes from Magic mirror in my hand, and we conduct experiments with changes based on it.

Usage

Next we describe how to use the code. The rest of the content and detailed will be gradually added to the project.

Generate Random Queries

The settings of randomly generated queries are listed in ARQG\settings.py, including the following content:

• the number of queries
• the length of the sql
• the database schema ( table name, join key and attributes )

Run ARQG\main.py to get the randomly generated queries, and the results are stored in ARQG\rand.sql.

Train the ChangeFormer

The ChangeFormer training code is located in index_selection_evaluation/ChangeFormer/trainer.py.

Running index selection

python -m selection