metadata
dataset_info:
features:
- name: images
sequence: image
- name: texts
list:
- name: user
dtype: string
- name: assistant
dtype: string
- name: source
dtype: string
splits:
- name: train
num_bytes: 244951255658.16818
num_examples: 565009
download_size: 145422811605
dataset_size: 244951255658.16818
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
Docmatix, but with multi-image samples filtered out.