metadata
dataset_info:
features:
- name: text
dtype: string
- name: token_idx
dtype: int64
- name: hidden_states
sequence: float64
splits:
- name: train
configs:
- config_name: default
data_files:
- split: train
path:
- c4/data_*
- cosmopedia_web/data_*
- redpajama_cc/data_*