metadata
dataset_info:
features:
- name: conversations
sequence:
- name: value
dtype: string
- name: source
dtype: string
- name: chosen
struct:
- name: value
dtype: string
- name: source
dtype: string
- name: rejected
struct:
- name: value
dtype: string
- name: source
dtype: string
- name: images
sequence: image
- name: task_id
dtype: int32
- name: step_id
dtype: int32
- name: source_name
dtype: string
splits:
- name: train
num_bytes: 9761332015.112
num_examples: 30921
download_size: 2962546411
dataset_size: 9761332015.112
configs:
- config_name: default
data_files:
- split: train
path: data/train-*