mkll/explain-extended-2024-gpt-sql

GPT in 500 lines of SQL

Explain Extended New Year 2024: GPT in 500 lines of SQL

https://explainextended.com/2023/12/31/happy-new-year-15/

About

This repository contains the code for the article Explain Extended New Year 2024: GPT in 500 lines of SQL.

The SQL code described in the article runs on PostgreSQL 15.0 and uses pgvector. It depends on tables populated with the parameters of the GPT-2 language model, which are not included in this repository.

To create the database and the tables, you need to build and run the Docker image explainextended2024. Alternatively, you can manually create the database, install pgvector, and run the script create.sql.

To populate the tables, you need to run the script populate.py. This script will download the GPT-2 parameters from OpenAI's website and populate the tables.

Requirements

• Docker
• Python 3.8 or later

Installation

1. Build and run the database image:

   docker build --tag explainextended2024 .
   docker run -d --name pgvector -e POSTGRES_DB=explainextended2024 -e POSTGRES_PASSWORD=MySecretPassword -e POSTGRES_USER=explainextended2024 -p 5432:5432 -v "explainextended2024:/var/lib/postgresql/data" explainextended2024
   

   Consider changing the password if your machine is exposed to the Internet.

2. Install Python dependencies:

   pip install -r requirements.txt
   

3. Populate the tables:

   python populate.py --connection_string="postgresql://explainextended2024:MySecretPassword@localhost:5432/explainextended2024"
   

   Make sure you're using the same password as in the docker run command.

4. The scripts mentioned in the article are located in the scripts directory.

Enjoy!