metadata
dataset_info:
features:
- name: chosen
list:
- name: content
dtype: string
- name: role
dtype: string
- name: reject
list:
- name: content
dtype: string
- name: role
dtype: string
- name: chosen_token
sequence: int64
- name: reject_token
sequence: int64
- name: chosen_mask
sequence: int64
- name: reject_mask
sequence: int64
- name: chosen_reward
dtype: float64
- name: reject_reward
dtype: float64
- name: chosen_logprob
dtype: float64
- name: reject_logprob
dtype: float64
splits:
- name: train
num_bytes: 4551323699
num_examples: 64101
- name: test
num_bytes: 35548914
num_examples: 500
download_size: 272310484
dataset_size: 4586872613
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*