metadata
dataset_info:
features:
- name: input_ids
sequence: int32
- name: token_type_ids
sequence: int8
- name: attention_mask
sequence: int8
- name: overflow_to_sample_mapping
dtype: int64
splits:
- name: train
num_bytes: 5971124282
num_examples: 6958840
download_size: 1363371505
dataset_size: 5971124282
configs:
- config_name: default
data_files:
- split: train
path: data/train-*