metadata
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*
dataset_info:
features:
- name: id
dtype: string
- name: title
dtype: string
- name: document
dtype: string
- name: question
dtype: string
- name: long_answers
sequence: string
- name: short_answers
sequence: string
splits:
- name: train
num_bytes: 4346873866.211105
num_examples: 106926
- name: validation
num_bytes: 175230324.62247765
num_examples: 4289
download_size: 1406784865
dataset_size: 4522104190.833583
Dataset Card for "natural_questions_clean"
Created by @thomwolf on the basis of https://huggingface.co/datasets/lighteval/natural_questions but removing the questions without short answers provided.