metadata
dataset_info:
features:
- name: input_ids
sequence: int32
- name: attention_mask
sequence: int8
- name: bbox
sequence:
sequence: int64
- name: start_positions
dtype: int64
- name: end_positions
dtype: int64
splits:
- name: train
num_bytes: 1693749560
num_examples: 80578
download_size: 77099140
dataset_size: 1693749560
configs:
- config_name: default
data_files:
- split: train
path: data/train-*