metadata
dataset_info:
features:
- name: token_id
dtype: int64
- name: data_id
dtype: int64
- name: divergent
dtype: float32
- name: small_token
dtype: int64
- name: real_token
dtype: int64
- name: small_logits
list: float32
- name: small_indices
list: int64
- name: small_last_hidden_states
list: float32
- name: mismatch
dtype: float32
- name: mask
dtype: float32
splits:
- name: train
num_bytes: 59200000
num_examples: 8000
download_size: 27525613
dataset_size: 59200000
configs:
- config_name: default
data_files:
- split: train
path: data/train-*