metadata
dataset_info:
features:
- name: input_ids
sequence: int32
- name: token_type_ids
sequence: int8
- name: attention_mask
sequence: int8
- name: pixel_values
sequence:
sequence:
sequence: float32
- name: pixel_mask
sequence:
sequence: int64
- name: labels
sequence: float32
splits:
- name: train
num_bytes: 21233639344
num_examples: 3914
download_size: 938336420
dataset_size: 21233639344
configs:
- config_name: default
data_files:
- split: train
path: data/train-*