metadata
dataset_info:
features:
- name: input_ids
sequence: int32
- name: attention_mask
sequence: int8
- name: labels
sequence: int64
- name: global_attention_mask
sequence: int64
- name: dataset_index
dtype: int64
splits:
- name: train
num_bytes: 5096195263
num_examples: 58195
download_size: 552583139
dataset_size: 5096195263
configs:
- config_name: default
data_files:
- split: train
path: data/train-*