metadata
dataset_info:
features:
- name: id
dtype: int64
- name: latent
dtype: int64
- name: layer
dtype: int64
- name: token_id
dtype: int64
- name: token
dtype: string
- name: act
dtype: float64
- name: token_ids
dtype: string
- name: tokens
dtype: string
- name: acts
dtype: string
splits:
- name: train
num_bytes: 1366975919
num_examples: 5778895
download_size: 227663486
dataset_size: 1366975919
configs:
- config_name: default
data_files:
- split: train
path: data/train-*