metadata
dataset_info:
features:
- name: input_ids
sequence: int32
- name: attention_mask
sequence: int8
- name: labels
sequence: int64
- name: sentence_start_positions
sequence: int64
splits:
- name: train
num_bytes: 6953309104
num_examples: 16000
download_size: 2203953540
dataset_size: 6953309104
configs:
- config_name: default
data_files:
- split: train
path: data/train-*