metadata
dataset_info:
- config_name: aggregation
features:
- name: sentence1
dtype: string
- name: labels
dtype: string
- name: metadata
dtype: string
splits:
- name: train
num_bytes: 5349271
num_examples: 4194
- name: validation
num_bytes: 614346
num_examples: 611
download_size: 1009796
dataset_size: 5963617
- config_name: column
features:
- name: inputs
dtype: string
- name: choices_list
list: string
- name: labels
dtype: int64
- name: metadata
dtype: string
splits:
- name: train
num_bytes: 5125886
num_examples: 2624
- name: validation
num_bytes: 505191
num_examples: 326
download_size: 2729701
dataset_size: 5631077
configs:
- config_name: aggregation
data_files:
- split: train
path: aggregation/train-*
- split: validation
path: aggregation/validation-*
- config_name: column
data_files:
- split: train
path: column/train-*
- split: validation
path: column/validation-*
license: cc-by-sa-4.0
source_datasets:
- minktn/spider-data
spider
Spider original parsed SQL and database schemas, single-output decision views.
Source: https://huggingface.co/datasets/minktn/spider-data, original distribution https://github.com/taoyds/spider. Native 7,000-question train and database-disjoint dev are retained. No hidden test or WikiSQL-derived train_others is included. Gold SQL never enters inputs. Arithmetic/set operations, multi-output queries and foreign-key-equivalent target columns are excluded. CC BY-SA 4.0.
Original data: minktn/spider-data, original distribution. Repackaged as parquet for tasksource by scripts/repackage_dataset/linguistic.py.
Conversion provenance: source revisions, hashes, configuration and exclusions.