metadata
dataset_info:
features:
- name: input_ids
sequence: int32
- name: attention_mask
sequence: int32
- name: token_type_ids
sequence: int32
- name: labels
sequence: int32
- name: pixel_values
sequence:
sequence:
sequence: float32
- name: masked_indices
sequence: int32
- name: metadata
struct:
- name: chunk_index
dtype: int64
- name: include_only_masked_tokens_images
dtype: bool
- name: logic_name
dtype: string
- name: mask_images_equally
dtype: bool
splits:
- name: train
num_bytes: 10901676897.960926
num_examples: 6142
- name: test
num_bytes: 2726306688.107173
num_examples: 1536
download_size: 3002617
dataset_size: 13627983586.0681
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*