metadata
dataset_info:
features:
- name: original_idx
dtype: int64
- name: input_ids
sequence: int32
- name: attention_mask
sequence: int8
- name: labels
sequence: int64
splits:
- name: train
num_bytes: 60117511545.66621
num_examples: 3352325
download_size: 2709527587
dataset_size: 60117511545.66621
configs:
- config_name: default
data_files:
- split: train
path: data/tokenized_with_idx/train-*