metadata
dataset_info:
features:
- name: id
dtype: string
- name: conversations
list:
- name: content
dtype: string
- name: role
dtype: string
- name: metadata
struct:
- name: any_step_biased
dtype: bool
- name: completion_idx
dtype: int64
- name: config
dtype: string
- name: conversation_reward
dtype: string
- name: example_id
dtype: int64
splits:
- name: train
num_bytes: 104604360
num_examples: 78954
download_size: 23510936
dataset_size: 104604360
configs:
- config_name: default
data_files:
- split: train
path: data/train-*