metadata
dataset_info:
features:
- name: text
dtype: string
- name: token_idx
dtype: int64
- name: hidden_states
sequence: float64
splits:
- name: train
num_bytes: 45948346550
num_examples: 2500100
download_size: 32233584920
dataset_size: 45948346550
configs:
- config_name: default
data_files:
- split: train
path: 400_499/train-*