metadata
license: cc-by-nc-sa-4.0
configs:
- config_name: aug
data_files:
- split: train
path: data/aug/*.parquet
- config_name: gen
data_files:
- split: train
path: data/gen/*.parquet
DataFlow-SQL-150k
DataFlow-SQL-150k is a text-to-SQL dataset containing 150,764 examples across two subsets: Aug and Gen. Both subsets provide natural-language questions, SQL queries, and generated reasoning text for text-to-SQL research.
Subsets
| Subset | Description | Examples |
|---|---|---|
| Aug | SQL-based data augmentation on Spider, BIRD, and EHRSQL. | 75,386 |
| Gen | Generated text-to-SQL examples with SQL complexity and question style labels. | 75,378 |
| Total | 150,764 |
Usage
from datasets import load_dataset
repo_id = "debugger123/DataFlow-SQL-150k"
aug = load_dataset(repo_id, "aug", split="train")
gen = load_dataset(repo_id, "gen", split="train")
The SQL and reasoning fields are SQL / cot_reasoning in Aug and sql / cot in Gen. Original field names are preserved.