metadata
dataset_info:
features:
- name: output
dtype: string
- name: input
dtype: string
- name: instruction
dtype: string
- name: activations
sequence: float64
splits:
- name: train
num_bytes: 162863372226
num_examples: 51760
download_size: 118635094090
dataset_size: 162863372226
configs:
- config_name: default
data_files:
- split: train
path: data/train-*