Open-source SQLCoder2 Large Language Model - Free deployment, easily convert natural language to SQL queries

Sqlcoder2

Developed by defog

Defog's SQLCoder is a cutting-edge large language model specializing in converting natural language questions into SQL queries.

Large Language Model

Transformers

Supports Multiple LanguagesOpen Source License:Other #Natural Language to SQL #Database Query Optimization #Surpassing GPT-3.5

Downloads 528

Release Time : 10/2/2023

Model Overview

SQLCoder is a 15-billion-parameter model that surpasses gpt-3.5-turbo in natural language to SQL performance and significantly outperforms all mainstream open-source models. When fine-tuned for specific database schemas, its performance even exceeds that of gpt-4.

Model Features

High-Performance SQL Generation

Outperforms gpt-3.5-turbo in the sql-eval evaluation framework and significantly surpasses mainstream open-source models.

Specialized Database Support

When fine-tuned for specific database schemas, it performs better than gpt-4.

Extensive Training Data

Trained on over 20,000 human-annotated questions covering 10 different database schemas.

Model Capabilities

Natural Language to SQL

Database Query Generation

Complex SQL Statement Generation

Use Cases

Data Analysis

Business Intelligence Analysis

Automatically converts non-technical users' natural language questions into SQL queries for business data analysis.

Improves data access efficiency for non-technical users.

Database Management

Helps database administrators quickly generate complex SQL queries.

Reduces SQL writing time.

🚀 Defog SQLCoder

Defog's SQLCoder is a cutting-edge large language model (LLM) designed to convert natural language questions into SQL queries, offering a seamless solution for data interaction.

Interactive Demo | 🤗 HF Repo | ♾️ Colab | 🐦 Twitter

🚀 Quick Start

SQLCoder is a 15B parameter model. On our sql-eval framework, it outperforms gpt-3.5-turbo in natural language to SQL generation tasks and significantly surpasses all popular open - source models. When fine - tuned on a given schema, it even outperforms gpt-4. It is fine - tuned on a base StarCoder model.

✨ Features

Results on novel datasets not seen in training

model	perc_correct
gpt4-2023-10-04	82.0
defog-sqlcoder2	77.5
gpt4-2023-08-28	74.0
defog-sqlcoder-7b	71.0
gpt-3.5-2023-10-04	66.0
claude-2	64.5
gpt-3.5-2023-08-28	61.0
claude_instant_1	61.0
text-davinci-003	52.5

Results by question category

We classified each generated question into one of 5 categories. The table shows the percentage of questions answered correctly by each model, broken down by category.

query_category	gpt-4	sqlcoder2-15b	sqlcoder-7b	gpt-3.5	claude-2	claude-instant	gpt-3
date	72	80	64	68	52	48	32
group_by	91.4	82.9	82.9	77.1	71.4	71.4	71.4
order_by	82.9	77.1	74.3	68.6	74.3	74.3	68.6
ratio	80	74.3	54.3	37.1	57.1	45.7	25.7
join	82.9	74.3	74.3	71.4	65.7	62.9	57.1
where	80	77.1	74.3	74.3	62.9	60	54.3

📦 Installation

You can use SQLCoder via the transformers library by downloading our model weights from the Hugging Face repo. We have added sample code for inference on a sample database schema.

python inference.py -q "Question about the sample database goes here"

# Sample question:
# Do we get more revenue from customers in New York compared to customers in San Francisco? Give me the total revenue for each city, and the difference between the two.

You can also use a demo on our website here, or run SQLCoder in Colab here

📚 Documentation

Training

Defog was trained on more than 20,000 human - curated questions. These questions were based on 10 different schemas. None of the schemas in the training data were included in our evaluation framework.

You can read more about our training approach and evaluation framework.

Hardware Requirements

SQLCoder has been tested on an A100 40GB GPU with bfloat16 weights. You can also load an 8 - bit and 4 - bit quantized version of the model on consumer GPUs with 20GB or more of memory – like RTX 4090, RTX 3090, and Apple M2 Pro, M2 Max, or M2 Ultra Chips with 20GB or more of memory.

📄 License

The code in this repo (what little there is of it) is Apache - 2 licensed. The model weights have a CC BY - SA 4.0 license, with additional responsible use restrictions added. The TL;DR is that you can use and modify the model for any purpose – including commercial use. However, if you modify the weights (for example, by fine - tuning), you must open - source your modified weights under the same license terms.

🔜 Todo

[x] Open - source the v1 model weights
[x] Train the model on more data, with higher data variance
[ ] Tune the model further with Reward Modelling and RLHF
[ ] Pretrain a model from scratch that specializes in SQL analysis

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご