metadata
language:
- en
license: mit
size_categories:
- 1M<n<10M
task_categories:
- text-generation
pretty_name: CLIP and T5 tokenization of CC3M
dataset_info:
features:
- name: text
dtype: string
- name: clip_ids
sequence: int64
- name: clip_attention_mask
sequence: int64
- name: t5_ids
sequence: int64
- name: t5_attention_mask
sequence: int64
splits:
- name: train
num_bytes: 31520132297
num_examples: 3318333
- name: validation
num_bytes: 150459428
num_examples: 15840
download_size: 362821979
dataset_size: 31670591725
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*