Spaces:
Running
Running
tongyx361
commited on
Commit
β’
935b23c
0
Parent(s):
Initialize the demo.
Browse files- .gitattributes +34 -0
- README.md +13 -0
- app.py +241 -0
- requirements.txt +13 -0
.gitattributes
ADDED
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
*.7z filter=lfs diff=lfs merge=lfs -text
|
2 |
+
*.arrow filter=lfs diff=lfs merge=lfs -text
|
3 |
+
*.bin filter=lfs diff=lfs merge=lfs -text
|
4 |
+
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
5 |
+
*.ckpt filter=lfs diff=lfs merge=lfs -text
|
6 |
+
*.ftz filter=lfs diff=lfs merge=lfs -text
|
7 |
+
*.gz filter=lfs diff=lfs merge=lfs -text
|
8 |
+
*.h5 filter=lfs diff=lfs merge=lfs -text
|
9 |
+
*.joblib filter=lfs diff=lfs merge=lfs -text
|
10 |
+
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
11 |
+
*.mlmodel filter=lfs diff=lfs merge=lfs -text
|
12 |
+
*.model filter=lfs diff=lfs merge=lfs -text
|
13 |
+
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
14 |
+
*.npy filter=lfs diff=lfs merge=lfs -text
|
15 |
+
*.npz filter=lfs diff=lfs merge=lfs -text
|
16 |
+
*.onnx filter=lfs diff=lfs merge=lfs -text
|
17 |
+
*.ot filter=lfs diff=lfs merge=lfs -text
|
18 |
+
*.parquet filter=lfs diff=lfs merge=lfs -text
|
19 |
+
*.pb filter=lfs diff=lfs merge=lfs -text
|
20 |
+
*.pickle filter=lfs diff=lfs merge=lfs -text
|
21 |
+
*.pkl filter=lfs diff=lfs merge=lfs -text
|
22 |
+
*.pt filter=lfs diff=lfs merge=lfs -text
|
23 |
+
*.pth filter=lfs diff=lfs merge=lfs -text
|
24 |
+
*.rar filter=lfs diff=lfs merge=lfs -text
|
25 |
+
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
26 |
+
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
27 |
+
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
28 |
+
*.tflite filter=lfs diff=lfs merge=lfs -text
|
29 |
+
*.tgz filter=lfs diff=lfs merge=lfs -text
|
30 |
+
*.wasm filter=lfs diff=lfs merge=lfs -text
|
31 |
+
*.xz filter=lfs diff=lfs merge=lfs -text
|
32 |
+
*.zip filter=lfs diff=lfs merge=lfs -text
|
33 |
+
*.zst filter=lfs diff=lfs merge=lfs -text
|
34 |
+
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
title: ImageReward Demo
|
3 |
+
emoji: π©βπ¨
|
4 |
+
colorFrom: indigo
|
5 |
+
colorTo: purple
|
6 |
+
sdk: gradio
|
7 |
+
sdk_version: 3.28.0
|
8 |
+
app_file: app.py
|
9 |
+
pinned: false
|
10 |
+
license: apache-2.0
|
11 |
+
---
|
12 |
+
|
13 |
+
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
app.py
ADDED
@@ -0,0 +1,241 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
|
3 |
+
import gradio as gr
|
4 |
+
import torch
|
5 |
+
from diffusers import StableDiffusionPipeline
|
6 |
+
from PIL import Image
|
7 |
+
|
8 |
+
import ImageReward as RM
|
9 |
+
|
10 |
+
# initialize
|
11 |
+
|
12 |
+
model_id = "runwayml/stable-diffusion-v1-5"
|
13 |
+
pipe = StableDiffusionPipeline.from_pretrained(
|
14 |
+
model_id,
|
15 |
+
torch_dtype=torch.float16,
|
16 |
+
)
|
17 |
+
|
18 |
+
model = RM.load("ImageReward-v1.0")
|
19 |
+
|
20 |
+
images_in_gallery = []
|
21 |
+
rewards_in_gallery = []
|
22 |
+
|
23 |
+
# event functions
|
24 |
+
|
25 |
+
|
26 |
+
def generate_images(
|
27 |
+
prompt, magic_words, num, height, width, num_inference_steps, guidance_scale
|
28 |
+
):
|
29 |
+
global images_in_gallery, rewards_in_gallery
|
30 |
+
|
31 |
+
if magic_words is not None:
|
32 |
+
prompt += ", ".join(magic_words)
|
33 |
+
|
34 |
+
images_in_gallery = pipe(
|
35 |
+
prompt,
|
36 |
+
height=height,
|
37 |
+
width=width,
|
38 |
+
num_inference_steps=num_inference_steps,
|
39 |
+
guidance_scale=guidance_scale,
|
40 |
+
num_images_per_prompt=num,
|
41 |
+
).images
|
42 |
+
rewards_in_gallery = [None] * len(images_in_gallery)
|
43 |
+
return list(zip(images_in_gallery, rewards_in_gallery))
|
44 |
+
|
45 |
+
|
46 |
+
def score_and_rank(prompt):
|
47 |
+
global rewards_in_gallery, images_in_gallery
|
48 |
+
|
49 |
+
num_not_scored = rewards_in_gallery.count(None)
|
50 |
+
|
51 |
+
if num_not_scored > 0:
|
52 |
+
images_to_score = images_in_gallery[-num_not_scored:]
|
53 |
+
with torch.no_grad():
|
54 |
+
ranks, rewards = model.inference_rank(prompt, images_to_score)
|
55 |
+
|
56 |
+
if not isinstance(rewards, list):
|
57 |
+
rewards = [rewards]
|
58 |
+
rewards_in_gallery = rewards_in_gallery[:-num_not_scored] + rewards
|
59 |
+
|
60 |
+
outputs = sorted(
|
61 |
+
zip(images_in_gallery, rewards_in_gallery), key=lambda x: x[1], reverse=True
|
62 |
+
)
|
63 |
+
|
64 |
+
images_in_gallery = [image for image, _ in outputs]
|
65 |
+
rewards_in_gallery = [reward for _, reward in outputs]
|
66 |
+
|
67 |
+
return outputs, [
|
68 |
+
[idx + 1, reward] for idx, reward in enumerate(rewards_in_gallery)
|
69 |
+
]
|
70 |
+
else:
|
71 |
+
return list(zip(images_in_gallery, rewards_in_gallery)), [
|
72 |
+
[idx + 1, reward] for idx, reward in enumerate(rewards_in_gallery)
|
73 |
+
]
|
74 |
+
|
75 |
+
|
76 |
+
def upload_images_to_gallery(uploaded_image_files):
|
77 |
+
global images_in_gallery, rewards_in_gallery
|
78 |
+
|
79 |
+
uploaded_image_file_paths = [file.name for file in uploaded_image_files]
|
80 |
+
uploaded_images = [Image.open(path) for path in uploaded_image_file_paths]
|
81 |
+
for path in uploaded_image_file_paths:
|
82 |
+
os.remove(path)
|
83 |
+
images_in_gallery = images_in_gallery + uploaded_images
|
84 |
+
rewards_in_gallery = rewards_in_gallery + [None] * len(uploaded_images)
|
85 |
+
|
86 |
+
return list(zip(images_in_gallery, rewards_in_gallery))
|
87 |
+
|
88 |
+
|
89 |
+
def clear_images():
|
90 |
+
global images_in_gallery, rewards_in_gallery
|
91 |
+
images_in_gallery = []
|
92 |
+
rewards_in_gallery = []
|
93 |
+
return None
|
94 |
+
|
95 |
+
|
96 |
+
if __name__ == "__main__":
|
97 |
+
# UI
|
98 |
+
with gr.Blocks(
|
99 |
+
theme=gr.themes.Monochrome(),
|
100 |
+
css=r".caption-label { color: black; }",
|
101 |
+
) as demo:
|
102 |
+
gr.HTML(
|
103 |
+
"""
|
104 |
+
<h1 align="center">ImageReward Demo</h1>
|
105 |
+
<p align="center"><a href="https://github.com/THUDM/ImageReward">GitHub Repo</a> β’ π€ <a href="https://huggingface.co/THUDM/ImageReward" target="_blank">HF Repo</a> β’ π¦ <a href="https://twitter.com/thukeg" target="_blank">Twitter</a> β’ π <a href="https://arxiv.org/abs/2304.05977" target="_blank">Paper</a><br></p>
|
106 |
+
<br>
|
107 |
+
<p dir="auto">ImageReward is the first general-purpose text-to-image <strong>human preference RM</strong>, which is trained on in total <strong>137k pairs of expert comparisons</strong>!</p>
|
108 |
+
<p dir="auto">The calculation of ImageRewards is based on <strong>both the prompt and images</strong>.</p>
|
109 |
+
"""
|
110 |
+
)
|
111 |
+
with gr.Row():
|
112 |
+
with gr.Column():
|
113 |
+
gr.HTML(
|
114 |
+
"""
|
115 |
+
<p dir="auto">Try ImageReward with only 2 steps:</p>
|
116 |
+
<ol dir="auto">
|
117 |
+
<li>Click the <strong>"Generate"</strong> button <strong>in the middle of the bottom</strong>.</li>
|
118 |
+
<li>Click the <strong>"Score&Rank"</strong> button <strong>below the gallery</strong>.</li>
|
119 |
+
</ol>
|
120 |
+
<p dir="auto">Finally, just check ImageRewards <strong>along with images or on the right of the gallery</strong>.</p>
|
121 |
+
<br>
|
122 |
+
<p dir="auto">This demo uses <code>runwayml/stable-diffusion-v1-5</code> as image generation model.</p>
|
123 |
+
"""
|
124 |
+
)
|
125 |
+
with gr.Column():
|
126 |
+
gr.HTML(
|
127 |
+
"""
|
128 |
+
<p dir="auto">Besides generating images, you can also <strong>upload</strong> images to score:</p>
|
129 |
+
<ol dir="auto">
|
130 |
+
<li>Upload images <strong>in the bottom right corner</strong>.</li>
|
131 |
+
<li>Change the <strong>"Prompt"</strong> to correspond to the images.</li>
|
132 |
+
<li>Click the <strong>"Score&Rank"</strong> button <strong>below the gallery</strong>.</li>
|
133 |
+
</ol>
|
134 |
+
<br>
|
135 |
+
<p dir="auto">For more details about using ImageReward in your own program, check <a href="https://github.com/THUDM/ImageReward">the README.md in our Github Repo</a>.</p>
|
136 |
+
"""
|
137 |
+
)
|
138 |
+
|
139 |
+
with gr.Row(elem_id="outputs_row"):
|
140 |
+
with gr.Column(elem_id="gallery_column", scale=4):
|
141 |
+
gallery = gr.Gallery(
|
142 |
+
label="Images (scored ones sorted)",
|
143 |
+
show_label=False,
|
144 |
+
elem_id="gallery",
|
145 |
+
).style(columns=4, object_fit="contain", full_width=True)
|
146 |
+
with gr.Column(elem_id="rewards_column"):
|
147 |
+
rewards = gr.Matrix(
|
148 |
+
value=[[None, None]],
|
149 |
+
headers=["Rank", "ImageReward"],
|
150 |
+
datatype="number",
|
151 |
+
)
|
152 |
+
with gr.Row():
|
153 |
+
score_and_rank_button = gr.Button("Score&Rank")
|
154 |
+
clear_button = gr.Button("Clear Gallery")
|
155 |
+
with gr.Row().style(equal_height=True):
|
156 |
+
with gr.Column():
|
157 |
+
prompt = gr.Textbox(
|
158 |
+
label="Prompt",
|
159 |
+
value="A painting of an ocean with clouds and birds, day time, low depth field effect, oil painting, impressionism",
|
160 |
+
)
|
161 |
+
|
162 |
+
examples = [
|
163 |
+
"A painting of an ocean with clouds and birds, day time, low depth field effect, oil painting, impressionism",
|
164 |
+
"A painting of a girl walking in a hallway and suddenly finds a giant sunflower on the floor blocking her way",
|
165 |
+
"Coronation of the sun emperor, digital art, illustration,4k resolution,intricate extremely detailed, depth,vivid colors",
|
166 |
+
"Symmetry!! Product render poster vivid colors divine proportion owl,glowing fog intricate,elegant, highly detailed",
|
167 |
+
"A unicorn in a clearing.it has a single shining horn. volumetric light.by emmanuel shiu, harry potter, eragon",
|
168 |
+
"Highly detailed portrait of a woman with long hairs,stephen bliss. unreal engine, fantasy art by greg rutkowski",
|
169 |
+
"Sculpture made of flame,portrait, female,future, torch,fire,harper's bazaar,vogue, fashion magazine, intricate",
|
170 |
+
]
|
171 |
+
prompt_examples = gr.Examples(
|
172 |
+
examples=examples,
|
173 |
+
label="Prompt Examples",
|
174 |
+
inputs=[prompt],
|
175 |
+
elem_id="prompt_examples",
|
176 |
+
)
|
177 |
+
|
178 |
+
with gr.Column():
|
179 |
+
choices = [
|
180 |
+
"HDR, UHD, 4K, 8K, 64K",
|
181 |
+
"highly detailed",
|
182 |
+
"studio lighting",
|
183 |
+
"professional",
|
184 |
+
"trending on artstation",
|
185 |
+
"unreal engine",
|
186 |
+
"vivid colors",
|
187 |
+
]
|
188 |
+
magic_words = gr.CheckboxGroup(
|
189 |
+
choices=choices,
|
190 |
+
value=choices,
|
191 |
+
type="value",
|
192 |
+
label="Magic Words to Append to Prompt",
|
193 |
+
)
|
194 |
+
|
195 |
+
num = gr.Slider(1, 16, step=1, label="Number of images", value=8)
|
196 |
+
height = gr.Slider(256, 2048, step=256, label="Height", value=512)
|
197 |
+
width = gr.Slider(256, 2048, step=256, label="Width", value=512)
|
198 |
+
num_inference_steps = gr.Slider(
|
199 |
+
0, 200, step=10, label="Number of inference steps", value=50
|
200 |
+
)
|
201 |
+
guidance_scale = gr.Slider(
|
202 |
+
0, 25, step=0.1, label="Guidance scale", value=7.5
|
203 |
+
)
|
204 |
+
|
205 |
+
generate_button = gr.Button("Generate")
|
206 |
+
with gr.Column():
|
207 |
+
gr.Markdown(
|
208 |
+
"""
|
209 |
+
- To clear all uploaded images, click the **"Clear Gallery"** button above.
|
210 |
+
- To clear the upload list and add additional images, click the **`x` in the upper right corner of the uploading window**.
|
211 |
+
- Additional images will be appended to the gallery, instead of replacing the existing ones.
|
212 |
+
"""
|
213 |
+
)
|
214 |
+
uploaded_image_files = gr.File(
|
215 |
+
file_count="multiple",
|
216 |
+
file_types=["image"],
|
217 |
+
type="file",
|
218 |
+
label="Upload Images",
|
219 |
+
show_label=True,
|
220 |
+
)
|
221 |
+
|
222 |
+
generate_button.click(
|
223 |
+
generate_images,
|
224 |
+
[
|
225 |
+
prompt,
|
226 |
+
magic_words,
|
227 |
+
num,
|
228 |
+
height,
|
229 |
+
width,
|
230 |
+
num_inference_steps,
|
231 |
+
guidance_scale,
|
232 |
+
],
|
233 |
+
[gallery],
|
234 |
+
)
|
235 |
+
score_and_rank_button.click(score_and_rank, [prompt], [gallery, rewards])
|
236 |
+
uploaded_image_files.upload(
|
237 |
+
upload_images_to_gallery, [uploaded_image_files], [gallery]
|
238 |
+
)
|
239 |
+
clear_button.click(clear_images, None, [gallery])
|
240 |
+
|
241 |
+
demo.launch()
|
requirements.txt
ADDED
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
image-reward
|
2 |
+
timm==0.6.13
|
3 |
+
transformers==4.27.4
|
4 |
+
fairscale==0.4.13
|
5 |
+
huggingface_hub==0.13.4
|
6 |
+
clip @ git+https://github.com/openai/CLIP.git
|
7 |
+
|
8 |
+
torch
|
9 |
+
diffusers
|
10 |
+
accelerate
|
11 |
+
|
12 |
+
Pillow
|
13 |
+
gradio
|