metadata
dataset_info:
features:
- name: image
dtype: image
- name: blip_caption_beam_5
dtype: string
- name: T5_last_hidden_states
sequence:
sequence:
sequence: float32
- name: sentences_raw
sequence: string
splits:
- name: train
num_bytes: 416620666
num_examples: 5000
download_size: 445433251
dataset_size: 416620666
configs:
- config_name: default
data_files:
- split: train
path: data/train-*