Skip to content

Dataset · gretelai

Synthetic Text-to-SQL dataset

gretelai/synthetic_text_to_sql

105k synthetic pairs for translating natural language questions into SQL queries.

What's inside

The dataset contains 105,851 synthetic Text-to-SQL examples: 100k records in train and 5,851 in test. The volume is about 23 million tokens, including roughly 12 million SQL tokens. The English data was generated via Gretel Navigator and published under Apache 2.0.

What you can do with it

  • Training Text-to-SQL models
  • Fine-tuning LLMs for table question answering
  • Testing SQL generation from questions
Tasks
Question answering, Table question answering, Text generation
Languages
English
Size
100K–1M rows
Downloads this month
6.7K
Likes
721

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("gretelai/synthetic_text_to_sql")