metadata
dataset_info:
features:
- name: caption
dtype: string
- name: image_tokens
dtype: string
- name: user
dtype: string
- name: assistant
dtype: string
- name: conversations
list:
- name: content
dtype: string
- name: role
dtype: string
splits:
- name: train
num_bytes: 10900258198
num_examples: 519288
download_size: 4159919663
dataset_size: 10900258198
configs:
- config_name: default
data_files:
- split: train
path: data/train-*