metadata
dataset_info:
features:
- name: doc_id
dtype: string
- name: meta
dtype: string
- name: quality_signals
dtype: string
- name: domain
dtype: string
- name: input_ids
sequence: int32
- name: attention_mask
sequence: int8
- name: special_tokens_mask
sequence: int8
- name: token_count
dtype: int64
splits:
- name: train
num_bytes: 67065187806
num_examples: 4156986
download_size: 21172612631
dataset_size: 67065187806
configs:
- config_name: default
data_files:
- split: train
path: data/train-*