metadata
dataset_info:
features:
- name: prompt
list:
- name: content
dtype: string
- name: role
dtype: string
- name: answer
dtype: string
- name: old_rewards
dtype: float32
splits:
- name: train
num_bytes: 555080
num_examples: 1494
download_size: 252832
dataset_size: 555080
configs:
- config_name: default
data_files:
- split: train
path: data/train-*