metadata
dataset_info:
features:
- name: image
dtype: image
- name: ground_truth
dtype: string
- name: labels
dtype: string
splits:
- name: train
num_bytes: 339139900
num_examples: 4000
download_size: 332704349
dataset_size: 339139900
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
deepseek ocr labeled