metadata
dataset_info:
features:
- name: original_image
dtype: image
- name: edited_image
dtype: image
- name: instruction
dtype: string
- name: original_instruction
dtype: string
- name: instruction_source
dtype: string
- name: clip_d_improvement
dtype: float32
- name: clip_d_old
dtype: float32
- name: clip_d_new
dtype: float32
- name: pickscore_improvement
dtype: float32
- name: clip_t_improvement
dtype: float32
splits:
- name: train
num_bytes: 3464233866
num_examples: 1000
download_size: 3464553839
dataset_size: 3464233866
configs:
- config_name: default
data_files:
- split: train
path: data/train-*