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: 6280833850
num_examples: 7881542
download_size: 1475190014
dataset_size: 6280833850
configs:
- config_name: default
data_files:
- split: train
path: data/train-*