motherduckdb/duckdb-text2sql-25k

---
license: cc-by-sa-4.0
task_categories:
- text2text-generation
language:
- en
tags:
- text-2-sql
pretty_name: duckdb-text2sql-25k
size_categories:
- 10K<n<100K
---

# Dataset Summary

The duckdb-text2sql-25k dataset contains 25,000 DuckDB text-2-sql pairs covering diverse aspects of DuckDB's SQL syntax.

We synthesized this dataset using [Mixtral 8x7B](https://huggingface.co/mistralai/Mixtral-8x7B-v0.1), based on [DuckDB's v0.9.2 documentation](https://duckdb.org/docs/archive/0.9/) and [Spider](https://huggingface.co/datasets/spider) schemas that were translated to DuckDB syntax and enriched with nested type columns.

Each training sample consists of a natural language prompt, a corresponding (optional) schema, and a resulting query. Each pair furthermore has a category property that indicates which part of the documentation was used generate the sample.

We applied various techniques to validate the syntactical and semantic correctness of the synthesized statements.

# How to use it

```python
from datasets import load_dataset

dataset = load_dataset("motherduckdb/duckdb-text2sql-25k")
```

We recommend using a prompt template similar to the one used for [DuckDB-NSQL-7B](https://huggingface.co/motherduckdb/DuckDB-NSQL-7B-v0.1#how-to-use) training.

# Dataset Structure

## Data Fields

- `prompt` (string): the instruction to generate SQL.
- `query` (string): the SQL statement.
- `schema` (string): the associated schema as CREATE TABLE statements.
- `category` (string): the category of the query.

# Languages

The language of the data is primarily English.

# Source Data and Licensing Information

Schemas in this dataset are derived from [Spider](https://huggingface.co/datasets/spider), with CC-BY-SA-4.0 License. We publish our dataset under the same license.

原始信息汇总

数据集概述

数据集名称

• pretty_name: duckdb-text2sql-25k

数据集内容

• 描述: 包含25,000个DuckDB文本到SQL的配对，覆盖DuckDB SQL语法的多个方面。
• 生成方式: 使用Mixtral 8x7B模型合成，基于DuckDB v0.9.2文档和Spider数据集的架构，经过转换和增强处理。
• 样本组成: 每个样本包括自然语言提示、相应的（可选）架构和生成的查询，以及指示生成样本所用文档部分的类别属性。

数据集结构

• 数据字段:
  • prompt (字符串): 生成SQL的指令。
  • query (字符串): SQL语句。
  • schema (字符串): 相关的架构，以CREATE TABLE语句表示。
  • category (字符串): 查询的类别。

语言

• 主要语言: 英语

许可信息

• 许可证: CC-BY-SA-4.0
• 源数据许可: 架构来源于Spider数据集，同样使用CC-BY-SA-4.0许可证。