metadata
dataset_info:
features:
- name: refs
sequence: string
- name: cand
dtype: string
- name: img
sequence:
sequence:
sequence: float32
- name: human_score
sequence: float32
splits:
- name: train
num_bytes: 308823663543
num_examples: 81909
- name: validation
num_bytes: 108781305987
num_examples: 28852
- name: test
num_bytes: 108743395430
num_examples: 28842
download_size: 122109791769
dataset_size: 526348364960
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*
- split: test
path: data/test-*