metadata
dataset_info:
features:
- name: id
dtype: string
- name: prompt
dtype: string
- name: latents
dtype:
array4_d:
shape:
- 2
- 4
- 128
- 128
dtype: float32
- name: prompt_embed
dtype:
array3_d:
shape:
- 1
- 77
- 2048
dtype: float32
- name: pooled_prompt_embed
dtype:
array2_d:
shape:
- 1
- 1280
dtype: float32
splits:
- name: train
num_bytes: 46590952683
num_examples: 40000
download_size: 21828419771
dataset_size: 46590952683
configs:
- config_name: default
data_files:
- split: train
path: data/train-*