Spaces:
Running
on
Zero
Running
on
Zero
Stalbe-X
commited on
Commit
·
b89eee2
0
Parent(s):
Initial Commit
Browse files- .gitattributes +41 -0
- .gitignore +4 -0
- README.md +16 -0
- app.py +506 -0
- requirements.txt +132 -0
- requirements_min.txt +17 -0
- setup.py +9 -0
- stablenormal/__init__.py +0 -0
- stablenormal/pipeline_stablenormal.py +1201 -0
- stablenormal/pipeline_yoso_normal.py +723 -0
- stablenormal/scheduler/__init__.py +0 -0
- stablenormal/stablecontrolnet.py +1354 -0
.gitattributes
ADDED
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
+
*.tar filter=lfs diff=lfs merge=lfs -text
|
29 |
+
*.tflite filter=lfs diff=lfs merge=lfs -text
|
30 |
+
*.tgz filter=lfs diff=lfs merge=lfs -text
|
31 |
+
*.wasm filter=lfs diff=lfs merge=lfs -text
|
32 |
+
*.xz filter=lfs diff=lfs merge=lfs -text
|
33 |
+
*.zip filter=lfs diff=lfs merge=lfs -text
|
34 |
+
*.zst filter=lfs diff=lfs merge=lfs -text
|
35 |
+
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
36 |
+
*.stl filter=lfs diff=lfs merge=lfs -text
|
37 |
+
*.glb filter=lfs diff=lfs merge=lfs -text
|
38 |
+
*.jpg filter=lfs diff=lfs merge=lfs -text
|
39 |
+
*.jpeg filter=lfs diff=lfs merge=lfs -text
|
40 |
+
*.png filter=lfs diff=lfs merge=lfs -text
|
41 |
+
*.mp4 filter=lfs diff=lfs merge=lfs -text
|
.gitignore
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
.idea
|
2 |
+
.DS_Store
|
3 |
+
__pycache__
|
4 |
+
weights
|
README.md
ADDED
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
title: Stable Normal Estimation
|
3 |
+
emoji: 🏵️
|
4 |
+
colorFrom: blue
|
5 |
+
colorTo: red
|
6 |
+
sdk: gradio
|
7 |
+
sdk_version: 4.32.2
|
8 |
+
app_file: app.py
|
9 |
+
pinned: true
|
10 |
+
license: cc-by-sa-4.0
|
11 |
+
models:
|
12 |
+
- Stable-X/stable-normal-v0-1
|
13 |
+
- Stable-X/yoso-normal-v0-1
|
14 |
+
hf_oauth: true
|
15 |
+
hf_oauth_expiration_minutes: 43200
|
16 |
+
---
|
app.py
ADDED
@@ -0,0 +1,506 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Copyright 2024 Anton Obukhov, ETH Zurich. All rights reserved.
|
2 |
+
#
|
3 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
4 |
+
# you may not use this file except in compliance with the License.
|
5 |
+
# You may obtain a copy of the License at
|
6 |
+
#
|
7 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
8 |
+
#
|
9 |
+
# Unless required by applicable law or agreed to in writing, software
|
10 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
11 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
12 |
+
# See the License for the specific language governing permissions and
|
13 |
+
# limitations under the License.
|
14 |
+
# --------------------------------------------------------------------------
|
15 |
+
# If you find this code useful, we kindly ask you to cite our paper in your work.
|
16 |
+
# Please find bibtex at: https://github.com/prs-eth/Marigold#-citation
|
17 |
+
# More information about the method can be found at https://marigoldmonodepth.github.io
|
18 |
+
# --------------------------------------------------------------------------
|
19 |
+
from __future__ import annotations
|
20 |
+
|
21 |
+
import functools
|
22 |
+
import os
|
23 |
+
import tempfile
|
24 |
+
|
25 |
+
import diffusers
|
26 |
+
import gradio as gr
|
27 |
+
import imageio as imageio
|
28 |
+
import numpy as np
|
29 |
+
import spaces
|
30 |
+
import torch as torch
|
31 |
+
from PIL import Image
|
32 |
+
from gradio_imageslider import ImageSlider
|
33 |
+
from tqdm import tqdm
|
34 |
+
|
35 |
+
from pathlib import Path
|
36 |
+
|
37 |
+
import gradio
|
38 |
+
from gradio.utils import get_cache_folder
|
39 |
+
from stablenormal.pipeline_yoso_normal import YOSONormalsPipeline
|
40 |
+
from stablenormal.pipeline_stablenormal import StableNormalPipeline
|
41 |
+
from stablenormal.scheduler.heuristics_ddimsampler import HEURI_DDIMScheduler
|
42 |
+
|
43 |
+
class Examples(gradio.helpers.Examples):
|
44 |
+
def __init__(self, *args, directory_name=None, **kwargs):
|
45 |
+
super().__init__(*args, **kwargs, _initiated_directly=False)
|
46 |
+
if directory_name is not None:
|
47 |
+
self.cached_folder = get_cache_folder() / directory_name
|
48 |
+
self.cached_file = Path(self.cached_folder) / "log.csv"
|
49 |
+
self.create()
|
50 |
+
|
51 |
+
|
52 |
+
default_seed = 2024
|
53 |
+
default_batch_size = 1
|
54 |
+
|
55 |
+
default_image_processing_resolution = 768
|
56 |
+
|
57 |
+
default_video_num_inference_steps = 10
|
58 |
+
default_video_processing_resolution = 768
|
59 |
+
default_video_out_max_frames = 450
|
60 |
+
|
61 |
+
def process_image_check(path_input):
|
62 |
+
if path_input is None:
|
63 |
+
raise gr.Error(
|
64 |
+
"Missing image in the first pane: upload a file or use one from the gallery below."
|
65 |
+
)
|
66 |
+
|
67 |
+
def process_image(
|
68 |
+
pipe,
|
69 |
+
path_input,
|
70 |
+
):
|
71 |
+
name_base, name_ext = os.path.splitext(os.path.basename(path_input))
|
72 |
+
print(f"Processing image {name_base}{name_ext}")
|
73 |
+
|
74 |
+
path_output_dir = tempfile.mkdtemp()
|
75 |
+
# path_out_fp32 = os.path.join(path_output_dir, f"{name_base}_normal_fp32.npy")
|
76 |
+
path_out_png = os.path.join(path_output_dir, f"{name_base}_normal_colored.png")
|
77 |
+
|
78 |
+
input_image = Image.open(path_input)
|
79 |
+
input_image = center_crop(input_image)
|
80 |
+
|
81 |
+
pipe_out = pipe(
|
82 |
+
input_image,
|
83 |
+
match_input_resolution=False,
|
84 |
+
return_intermediate_result=False
|
85 |
+
)
|
86 |
+
|
87 |
+
normal_pred = pipe_out.prediction[0, :, :]
|
88 |
+
normal_colored = pipe.image_processor.visualize_normals(pipe_out.prediction)
|
89 |
+
normal_colored[-1].save(path_out_png)
|
90 |
+
print(path_out_png)
|
91 |
+
# np.save(path_out_fp32, normal_pred)
|
92 |
+
# path_out_vis = os.path.join(path_output_dir, f"{name_base}_normal_refinement_process.gif")
|
93 |
+
# normal_colored[0].save(path_out_vis, save_all=True,
|
94 |
+
# append_images=normal_colored[1:],
|
95 |
+
# duration=400, loop=0)
|
96 |
+
return [input_image, path_out_png]
|
97 |
+
|
98 |
+
def center_crop(img):
|
99 |
+
# Open the image file
|
100 |
+
img_width, img_height = img.size
|
101 |
+
crop_width =min(img_width, img_height)
|
102 |
+
# Calculate the cropping box
|
103 |
+
left = (img_width - crop_width) / 2
|
104 |
+
top = (img_height - crop_width) / 2
|
105 |
+
right = (img_width + crop_width) / 2
|
106 |
+
bottom = (img_height + crop_width) / 2
|
107 |
+
|
108 |
+
# Crop the image
|
109 |
+
img_cropped = img.crop((left, top, right, bottom))
|
110 |
+
return img_cropped
|
111 |
+
|
112 |
+
def process_video(
|
113 |
+
pipe,
|
114 |
+
path_input,
|
115 |
+
out_max_frames=default_video_out_max_frames,
|
116 |
+
target_fps=3,
|
117 |
+
progress=gr.Progress(),
|
118 |
+
):
|
119 |
+
if path_input is None:
|
120 |
+
raise gr.Error(
|
121 |
+
"Missing video in the first pane: upload a file or use one from the gallery below."
|
122 |
+
)
|
123 |
+
|
124 |
+
name_base, name_ext = os.path.splitext(os.path.basename(path_input))
|
125 |
+
print(f"Processing video {name_base}{name_ext}")
|
126 |
+
|
127 |
+
path_output_dir = tempfile.mkdtemp()
|
128 |
+
path_out_vis = os.path.join(path_output_dir, f"{name_base}_normal_colored.mp4")
|
129 |
+
|
130 |
+
reader, writer = None, None
|
131 |
+
try:
|
132 |
+
reader = imageio.get_reader(path_input)
|
133 |
+
|
134 |
+
meta_data = reader.get_meta_data()
|
135 |
+
fps = meta_data["fps"]
|
136 |
+
size = meta_data["size"]
|
137 |
+
duration_sec = meta_data["duration"]
|
138 |
+
|
139 |
+
writer = imageio.get_writer(path_out_vis, fps=target_fps)
|
140 |
+
|
141 |
+
out_frame_id = 0
|
142 |
+
pbar = tqdm(desc="Processing Video", total=duration_sec)
|
143 |
+
|
144 |
+
for frame_id, frame in enumerate(reader):
|
145 |
+
if frame_id % (fps // target_fps) != 0:
|
146 |
+
continue
|
147 |
+
else:
|
148 |
+
out_frame_id += 1
|
149 |
+
pbar.update(1)
|
150 |
+
if out_frame_id > out_max_frames:
|
151 |
+
break
|
152 |
+
|
153 |
+
frame_pil = Image.fromarray(frame)
|
154 |
+
frame_pil = center_crop(frame_pil)
|
155 |
+
pipe_out = pipe(
|
156 |
+
frame_pil,
|
157 |
+
match_input_resolution=False,
|
158 |
+
return_intermediate_result=False
|
159 |
+
)
|
160 |
+
|
161 |
+
processed_frame = pipe.image_processor.visualize_normals( # noqa
|
162 |
+
pipe_out.prediction
|
163 |
+
)[0]
|
164 |
+
processed_frame = np.array(processed_frame)
|
165 |
+
|
166 |
+
_processed_frame = imageio.core.util.Array(processed_frame)
|
167 |
+
writer.append_data(_processed_frame)
|
168 |
+
|
169 |
+
yield (
|
170 |
+
[frame_pil, processed_frame],
|
171 |
+
None,
|
172 |
+
)
|
173 |
+
finally:
|
174 |
+
|
175 |
+
if writer is not None:
|
176 |
+
writer.close()
|
177 |
+
|
178 |
+
if reader is not None:
|
179 |
+
reader.close()
|
180 |
+
|
181 |
+
yield (
|
182 |
+
[frame_pil, processed_frame],
|
183 |
+
[path_out_vis,]
|
184 |
+
)
|
185 |
+
|
186 |
+
|
187 |
+
def run_demo_server(pipe):
|
188 |
+
process_pipe_image = spaces.GPU(functools.partial(process_image, pipe))
|
189 |
+
process_pipe_video = spaces.GPU(
|
190 |
+
functools.partial(process_video, pipe), duration=120
|
191 |
+
)
|
192 |
+
|
193 |
+
gradio_theme = gr.themes.Default()
|
194 |
+
|
195 |
+
with gr.Blocks(
|
196 |
+
theme=gradio_theme,
|
197 |
+
title="Stable Normal Estimation",
|
198 |
+
css="""
|
199 |
+
#download {
|
200 |
+
height: 118px;
|
201 |
+
}
|
202 |
+
.slider .inner {
|
203 |
+
width: 5px;
|
204 |
+
background: #FFF;
|
205 |
+
}
|
206 |
+
.viewport {
|
207 |
+
aspect-ratio: 4/3;
|
208 |
+
}
|
209 |
+
.tabs button.selected {
|
210 |
+
font-size: 20px !important;
|
211 |
+
color: crimson !important;
|
212 |
+
}
|
213 |
+
h1 {
|
214 |
+
text-align: center;
|
215 |
+
display: block;
|
216 |
+
}
|
217 |
+
h2 {
|
218 |
+
text-align: center;
|
219 |
+
display: block;
|
220 |
+
}
|
221 |
+
h3 {
|
222 |
+
text-align: center;
|
223 |
+
display: block;
|
224 |
+
}
|
225 |
+
.md_feedback li {
|
226 |
+
margin-bottom: 0px !important;
|
227 |
+
}
|
228 |
+
""",
|
229 |
+
head="""
|
230 |
+
<script async src="https://www.googletagmanager.com/gtag/js?id=G-1FWSVCGZTG"></script>
|
231 |
+
<script>
|
232 |
+
window.dataLayer = window.dataLayer || [];
|
233 |
+
function gtag() {dataLayer.push(arguments);}
|
234 |
+
gtag('js', new Date());
|
235 |
+
gtag('config', 'G-1FWSVCGZTG');
|
236 |
+
</script>
|
237 |
+
""",
|
238 |
+
) as demo:
|
239 |
+
gr.Markdown(
|
240 |
+
"""
|
241 |
+
# StableNormal: Reducing Diffusion Variance for Stable and Sharp Normal
|
242 |
+
<p align="center">
|
243 |
+
"""
|
244 |
+
)
|
245 |
+
|
246 |
+
with gr.Tabs(elem_classes=["tabs"]):
|
247 |
+
with gr.Tab("Image"):
|
248 |
+
with gr.Row():
|
249 |
+
with gr.Column():
|
250 |
+
image_input = gr.Image(
|
251 |
+
label="Input Image",
|
252 |
+
type="filepath",
|
253 |
+
)
|
254 |
+
with gr.Row():
|
255 |
+
image_submit_btn = gr.Button(
|
256 |
+
value="Compute Normal", variant="primary"
|
257 |
+
)
|
258 |
+
image_reset_btn = gr.Button(value="Reset")
|
259 |
+
with gr.Column():
|
260 |
+
image_output_slider = ImageSlider(
|
261 |
+
label="Normal outputs",
|
262 |
+
type="filepath",
|
263 |
+
show_download_button=True,
|
264 |
+
show_share_button=True,
|
265 |
+
interactive=False,
|
266 |
+
elem_classes="slider",
|
267 |
+
position=0.25,
|
268 |
+
)
|
269 |
+
|
270 |
+
Examples(
|
271 |
+
fn=process_pipe_image,
|
272 |
+
examples=sorted([
|
273 |
+
os.path.join("files", "image", name)
|
274 |
+
for name in os.listdir(os.path.join("files", "image"))
|
275 |
+
]),
|
276 |
+
inputs=[image_input],
|
277 |
+
outputs=[image_output_slider],
|
278 |
+
cache_examples=True,
|
279 |
+
directory_name="examples_image",
|
280 |
+
)
|
281 |
+
|
282 |
+
with gr.Tab("Video"):
|
283 |
+
with gr.Row():
|
284 |
+
with gr.Column():
|
285 |
+
video_input = gr.Video(
|
286 |
+
label="Input Video",
|
287 |
+
sources=["upload", "webcam"],
|
288 |
+
)
|
289 |
+
with gr.Row():
|
290 |
+
video_submit_btn = gr.Button(
|
291 |
+
value="Compute Normal", variant="primary"
|
292 |
+
)
|
293 |
+
video_reset_btn = gr.Button(value="Reset")
|
294 |
+
with gr.Column():
|
295 |
+
processed_frames = ImageSlider(
|
296 |
+
label="Realtime Visualization",
|
297 |
+
type="filepath",
|
298 |
+
show_download_button=True,
|
299 |
+
show_share_button=True,
|
300 |
+
interactive=False,
|
301 |
+
elem_classes="slider",
|
302 |
+
position=0.25,
|
303 |
+
)
|
304 |
+
video_output_files = gr.Files(
|
305 |
+
label="Normal outputs",
|
306 |
+
elem_id="download",
|
307 |
+
interactive=False,
|
308 |
+
)
|
309 |
+
Examples(
|
310 |
+
fn=process_pipe_video,
|
311 |
+
examples=sorted([
|
312 |
+
os.path.join("files", "video", name)
|
313 |
+
for name in os.listdir(os.path.join("files", "video"))
|
314 |
+
]),
|
315 |
+
inputs=[video_input],
|
316 |
+
outputs=[processed_frames, video_output_files],
|
317 |
+
directory_name="examples_video",
|
318 |
+
cache_examples=True,
|
319 |
+
)
|
320 |
+
|
321 |
+
with gr.Tab("Panorama"):
|
322 |
+
with gr.Column():
|
323 |
+
gr.Markdown("Functionality coming soon on June.10th")
|
324 |
+
|
325 |
+
with gr.Tab("4K Image"):
|
326 |
+
with gr.Column():
|
327 |
+
gr.Markdown("Functionality coming soon on June.17th")
|
328 |
+
|
329 |
+
with gr.Tab("Normal Mapping"):
|
330 |
+
with gr.Column():
|
331 |
+
gr.Markdown("Functionality coming soon on June.24th")
|
332 |
+
|
333 |
+
with gr.Tab("Normal SuperResolution"):
|
334 |
+
with gr.Column():
|
335 |
+
gr.Markdown("Functionality coming soon on June.30th")
|
336 |
+
|
337 |
+
### Image tab
|
338 |
+
image_submit_btn.click(
|
339 |
+
fn=process_image_check,
|
340 |
+
inputs=image_input,
|
341 |
+
outputs=None,
|
342 |
+
preprocess=False,
|
343 |
+
queue=False,
|
344 |
+
).success(
|
345 |
+
fn=process_pipe_image,
|
346 |
+
inputs=[
|
347 |
+
image_input,
|
348 |
+
],
|
349 |
+
outputs=[image_output_slider],
|
350 |
+
concurrency_limit=1,
|
351 |
+
)
|
352 |
+
|
353 |
+
image_reset_btn.click(
|
354 |
+
fn=lambda: (
|
355 |
+
None,
|
356 |
+
None,
|
357 |
+
None,
|
358 |
+
),
|
359 |
+
inputs=[],
|
360 |
+
outputs=[
|
361 |
+
image_input,
|
362 |
+
image_output_slider,
|
363 |
+
],
|
364 |
+
queue=False,
|
365 |
+
)
|
366 |
+
|
367 |
+
### Video tab
|
368 |
+
|
369 |
+
video_submit_btn.click(
|
370 |
+
fn=process_pipe_video,
|
371 |
+
inputs=[video_input],
|
372 |
+
outputs=[processed_frames, video_output_files],
|
373 |
+
concurrency_limit=1,
|
374 |
+
)
|
375 |
+
|
376 |
+
video_reset_btn.click(
|
377 |
+
fn=lambda: (None, None, None),
|
378 |
+
inputs=[],
|
379 |
+
outputs=[video_input, processed_frames, video_output_files],
|
380 |
+
concurrency_limit=1,
|
381 |
+
)
|
382 |
+
|
383 |
+
### Server launch
|
384 |
+
|
385 |
+
demo.queue(
|
386 |
+
api_open=False,
|
387 |
+
).launch(
|
388 |
+
server_name="0.0.0.0",
|
389 |
+
server_port=7860,
|
390 |
+
)
|
391 |
+
|
392 |
+
from einops import rearrange
|
393 |
+
class DINOv2_Encoder:
|
394 |
+
IMAGENET_DEFAULT_MEAN = [0.485, 0.456, 0.406]
|
395 |
+
IMAGENET_DEFAULT_STD = [0.229, 0.224, 0.225]
|
396 |
+
|
397 |
+
def __init__(
|
398 |
+
self,
|
399 |
+
model_name = 'dinov2_vitl14',
|
400 |
+
freeze = True,
|
401 |
+
antialias=True,
|
402 |
+
device="cuda",
|
403 |
+
size = 448,
|
404 |
+
):
|
405 |
+
|
406 |
+
super(DINOv2_Encoder).__init__()
|
407 |
+
|
408 |
+
self.model = torch.hub.load('facebookresearch/dinov2', model_name)
|
409 |
+
self.model.eval()
|
410 |
+
self.device = device
|
411 |
+
self.antialias = antialias
|
412 |
+
self.dtype = torch.float32
|
413 |
+
|
414 |
+
self.mean = torch.Tensor(self.IMAGENET_DEFAULT_MEAN)
|
415 |
+
self.std = torch.Tensor(self.IMAGENET_DEFAULT_STD)
|
416 |
+
self.size = size
|
417 |
+
if freeze:
|
418 |
+
self.freeze()
|
419 |
+
|
420 |
+
|
421 |
+
def freeze(self):
|
422 |
+
for param in self.model.parameters():
|
423 |
+
param.requires_grad = False
|
424 |
+
|
425 |
+
@torch.no_grad()
|
426 |
+
def encoder(self, x):
|
427 |
+
'''
|
428 |
+
x: [b h w c], range from (-1, 1), rbg
|
429 |
+
'''
|
430 |
+
|
431 |
+
x = self.preprocess(x).to(self.device, self.dtype)
|
432 |
+
|
433 |
+
b, c, h, w = x.shape
|
434 |
+
patch_h, patch_w = h // 14, w // 14
|
435 |
+
|
436 |
+
embeddings = self.model.forward_features(x)['x_norm_patchtokens']
|
437 |
+
embeddings = rearrange(embeddings, 'b (h w) c -> b h w c', h = patch_h, w = patch_w)
|
438 |
+
|
439 |
+
return rearrange(embeddings, 'b h w c -> b c h w')
|
440 |
+
|
441 |
+
def preprocess(self, x):
|
442 |
+
''' x
|
443 |
+
'''
|
444 |
+
# normalize to [0,1],
|
445 |
+
x = torch.nn.functional.interpolate(
|
446 |
+
x,
|
447 |
+
size=(self.size, self.size),
|
448 |
+
mode='bicubic',
|
449 |
+
align_corners=True,
|
450 |
+
antialias=self.antialias,
|
451 |
+
)
|
452 |
+
|
453 |
+
x = (x + 1.0) / 2.0
|
454 |
+
# renormalize according to dino
|
455 |
+
mean = self.mean.view(1, 3, 1, 1).to(x.device)
|
456 |
+
std = self.std.view(1, 3, 1, 1).to(x.device)
|
457 |
+
x = (x - mean) / std
|
458 |
+
|
459 |
+
return x
|
460 |
+
|
461 |
+
def to(self, device, dtype=None):
|
462 |
+
if dtype is not None:
|
463 |
+
self.dtype = dtype
|
464 |
+
self.model.to(device, dtype)
|
465 |
+
self.mean.to(device, dtype)
|
466 |
+
self.std.to(device, dtype)
|
467 |
+
else:
|
468 |
+
self.model.to(device)
|
469 |
+
self.mean.to(device)
|
470 |
+
self.std.to(device)
|
471 |
+
return self
|
472 |
+
|
473 |
+
def __call__(self, x, **kwargs):
|
474 |
+
return self.encoder(x, **kwargs)
|
475 |
+
|
476 |
+
def main():
|
477 |
+
os.system("pip freeze")
|
478 |
+
|
479 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
480 |
+
|
481 |
+
x_start_pipeline = YOSONormalsPipeline.from_pretrained(
|
482 |
+
'weights/yoso-normal-v0-1', trust_remote_code=True,
|
483 |
+
t_start=300).to(device, torch.float16)
|
484 |
+
dinov2_prior = DINOv2_Encoder(size=672)
|
485 |
+
dinov2_prior.to(device, torch.float16)
|
486 |
+
|
487 |
+
pipe = StableNormalPipeline.from_pretrained('weights/stable-normal-v0-1', t_start=300, trust_remote_code=True,
|
488 |
+
scheduler=HEURI_DDIMScheduler(prediction_type='sample',
|
489 |
+
beta_start=0.00085, beta_end=0.0120,
|
490 |
+
beta_schedule = "scaled_linear"))
|
491 |
+
# two stage concat
|
492 |
+
pipe.x_start_pipeline = x_start_pipeline
|
493 |
+
pipe.prior = dinov2_prior
|
494 |
+
pipe.to(device, torch.float16)
|
495 |
+
|
496 |
+
try:
|
497 |
+
import xformers
|
498 |
+
pipe.enable_xformers_memory_efficient_attention()
|
499 |
+
except:
|
500 |
+
pass # run without xformers
|
501 |
+
|
502 |
+
run_demo_server(pipe)
|
503 |
+
|
504 |
+
|
505 |
+
if __name__ == "__main__":
|
506 |
+
main()
|
requirements.txt
ADDED
@@ -0,0 +1,132 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
accelerate==0.30.1
|
2 |
+
aiofiles==23.2.1
|
3 |
+
aiohttp==3.9.5
|
4 |
+
aiosignal==1.3.1
|
5 |
+
altair==5.3.0
|
6 |
+
annotated-types==0.7.0
|
7 |
+
anyio==4.4.0
|
8 |
+
async-timeout==4.0.3
|
9 |
+
attrs==23.2.0
|
10 |
+
Authlib==1.3.0
|
11 |
+
certifi==2024.2.2
|
12 |
+
cffi==1.16.0
|
13 |
+
charset-normalizer==3.3.2
|
14 |
+
click==8.0.4
|
15 |
+
contourpy==1.2.1
|
16 |
+
cryptography==42.0.7
|
17 |
+
cycler==0.12.1
|
18 |
+
dataclasses-json==0.6.6
|
19 |
+
datasets==2.19.1
|
20 |
+
Deprecated==1.2.14
|
21 |
+
diffusers==0.28.0
|
22 |
+
dill==0.3.8
|
23 |
+
dnspython==2.6.1
|
24 |
+
email_validator==2.1.1
|
25 |
+
exceptiongroup==1.2.1
|
26 |
+
fastapi==0.111.0
|
27 |
+
fastapi-cli==0.0.4
|
28 |
+
ffmpy==0.3.2
|
29 |
+
filelock==3.14.0
|
30 |
+
fonttools==4.53.0
|
31 |
+
frozenlist==1.4.1
|
32 |
+
fsspec==2024.3.1
|
33 |
+
gradio==4.32.2
|
34 |
+
gradio_client==0.17.0
|
35 |
+
gradio_imageslider==0.0.20
|
36 |
+
h11==0.14.0
|
37 |
+
httpcore==1.0.5
|
38 |
+
httptools==0.6.1
|
39 |
+
httpx==0.27.0
|
40 |
+
huggingface-hub==0.23.0
|
41 |
+
idna==3.7
|
42 |
+
imageio==2.34.1
|
43 |
+
imageio-ffmpeg==0.5.0
|
44 |
+
importlib_metadata==7.1.0
|
45 |
+
importlib_resources==6.4.0
|
46 |
+
itsdangerous==2.2.0
|
47 |
+
Jinja2==3.1.4
|
48 |
+
jsonschema==4.22.0
|
49 |
+
jsonschema-specifications==2023.12.1
|
50 |
+
kiwisolver==1.4.5
|
51 |
+
markdown-it-py==3.0.0
|
52 |
+
MarkupSafe==2.1.5
|
53 |
+
marshmallow==3.21.2
|
54 |
+
matplotlib==3.8.2
|
55 |
+
mdurl==0.1.2
|
56 |
+
mpmath==1.3.0
|
57 |
+
multidict==6.0.5
|
58 |
+
multiprocess==0.70.16
|
59 |
+
mypy-extensions==1.0.0
|
60 |
+
networkx==3.3
|
61 |
+
numpy==1.26.4
|
62 |
+
nvidia-cublas-cu12==12.1.3.1
|
63 |
+
nvidia-cuda-cupti-cu12==12.1.105
|
64 |
+
nvidia-cuda-nvrtc-cu12==12.1.105
|
65 |
+
nvidia-cuda-runtime-cu12==12.1.105
|
66 |
+
nvidia-cudnn-cu12==8.9.2.26
|
67 |
+
nvidia-cufft-cu12==11.0.2.54
|
68 |
+
nvidia-curand-cu12==10.3.2.106
|
69 |
+
nvidia-cusolver-cu12==11.4.5.107
|
70 |
+
nvidia-cusparse-cu12==12.1.0.106
|
71 |
+
nvidia-nccl-cu12==2.19.3
|
72 |
+
nvidia-nvjitlink-cu12==12.5.40
|
73 |
+
nvidia-nvtx-cu12==12.1.105
|
74 |
+
orjson==3.10.3
|
75 |
+
packaging==24.0
|
76 |
+
pandas==2.2.2
|
77 |
+
pillow==10.3.0
|
78 |
+
protobuf==3.20.3
|
79 |
+
psutil==5.9.8
|
80 |
+
pyarrow==16.0.0
|
81 |
+
pyarrow-hotfix==0.6
|
82 |
+
pycparser==2.22
|
83 |
+
pydantic==2.7.2
|
84 |
+
pydantic_core==2.18.3
|
85 |
+
pydub==0.25.1
|
86 |
+
pygltflib==1.16.1
|
87 |
+
Pygments==2.18.0
|
88 |
+
pyparsing==3.1.2
|
89 |
+
python-dateutil==2.9.0.post0
|
90 |
+
python-dotenv==1.0.1
|
91 |
+
python-multipart==0.0.9
|
92 |
+
pytz==2024.1
|
93 |
+
PyYAML==6.0.1
|
94 |
+
referencing==0.35.1
|
95 |
+
regex==2024.5.15
|
96 |
+
requests==2.31.0
|
97 |
+
rich==13.7.1
|
98 |
+
rpds-py==0.18.1
|
99 |
+
ruff==0.4.7
|
100 |
+
safetensors==0.4.3
|
101 |
+
scipy==1.11.4
|
102 |
+
semantic-version==2.10.0
|
103 |
+
shellingham==1.5.4
|
104 |
+
six==1.16.0
|
105 |
+
sniffio==1.3.1
|
106 |
+
spaces==0.28.3
|
107 |
+
starlette==0.37.2
|
108 |
+
sympy==1.12.1
|
109 |
+
tokenizers==0.15.2
|
110 |
+
tomlkit==0.12.0
|
111 |
+
toolz==0.12.1
|
112 |
+
torch==2.2.0
|
113 |
+
tqdm==4.66.4
|
114 |
+
transformers==4.36.1
|
115 |
+
trimesh==4.0.5
|
116 |
+
triton==2.2.0
|
117 |
+
typer==0.12.3
|
118 |
+
typing-inspect==0.9.0
|
119 |
+
typing_extensions==4.11.0
|
120 |
+
tzdata==2024.1
|
121 |
+
ujson==5.10.0
|
122 |
+
urllib3==2.2.1
|
123 |
+
uvicorn==0.30.0
|
124 |
+
uvloop==0.19.0
|
125 |
+
watchfiles==0.22.0
|
126 |
+
websockets==11.0.3
|
127 |
+
wrapt==1.16.0
|
128 |
+
xformers==0.0.24
|
129 |
+
xxhash==3.4.1
|
130 |
+
yarl==1.9.4
|
131 |
+
zipp==3.19.1
|
132 |
+
einops==0.7.0
|
requirements_min.txt
ADDED
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
gradio>=4.32.1
|
2 |
+
gradio-imageslider>=0.0.20
|
3 |
+
pygltflib==1.16.1
|
4 |
+
trimesh==4.0.5
|
5 |
+
imageio
|
6 |
+
imageio-ffmpeg
|
7 |
+
Pillow
|
8 |
+
einops==0.7.0
|
9 |
+
|
10 |
+
spaces
|
11 |
+
accelerate
|
12 |
+
diffusers>=0.28.0
|
13 |
+
matplotlib==3.8.2
|
14 |
+
scipy==1.11.4
|
15 |
+
torch==2.0.1
|
16 |
+
transformers==4.36.1
|
17 |
+
xformers==0.0.21
|
setup.py
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from pathlib import Path
|
2 |
+
from setuptools import setup, find_packages
|
3 |
+
|
4 |
+
setup_path = Path(__file__).parent
|
5 |
+
|
6 |
+
setup(
|
7 |
+
name = "stablenormal",
|
8 |
+
packages=find_packages()
|
9 |
+
)
|
stablenormal/__init__.py
ADDED
File without changes
|
stablenormal/pipeline_stablenormal.py
ADDED
@@ -0,0 +1,1201 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Copyright 2024 Marigold authors, PRS ETH Zurich. All rights reserved.
|
2 |
+
# Copyright 2024 The HuggingFace Team. All rights reserved.
|
3 |
+
#
|
4 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
5 |
+
# you may not use this file except in compliance with the License.
|
6 |
+
# You may obtain a copy of the License at
|
7 |
+
#
|
8 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
9 |
+
#
|
10 |
+
# Unless required by applicable law or agreed to in writing, software
|
11 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
12 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
13 |
+
# See the License for the specific language governing permissions and
|
14 |
+
# limitations under the License.
|
15 |
+
# --------------------------------------------------------------------------
|
16 |
+
# More information and citation instructions are available on the
|
17 |
+
# --------------------------------------------------------------------------
|
18 |
+
from dataclasses import dataclass
|
19 |
+
from typing import Any, Dict, List, Optional, Tuple, Union
|
20 |
+
|
21 |
+
import numpy as np
|
22 |
+
import torch
|
23 |
+
from PIL import Image
|
24 |
+
from tqdm.auto import tqdm
|
25 |
+
from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
|
26 |
+
|
27 |
+
|
28 |
+
from diffusers.image_processor import PipelineImageInput
|
29 |
+
from diffusers.models import (
|
30 |
+
AutoencoderKL,
|
31 |
+
UNet2DConditionModel,
|
32 |
+
ControlNetModel,
|
33 |
+
)
|
34 |
+
from diffusers.schedulers import (
|
35 |
+
DDIMScheduler
|
36 |
+
)
|
37 |
+
|
38 |
+
from diffusers.utils import (
|
39 |
+
BaseOutput,
|
40 |
+
logging,
|
41 |
+
replace_example_docstring,
|
42 |
+
)
|
43 |
+
|
44 |
+
from diffusers.models.unets.unet_2d_condition import UNet2DConditionOutput
|
45 |
+
|
46 |
+
from diffusers.utils import USE_PEFT_BACKEND, BaseOutput, deprecate, logging, scale_lora_layers, unscale_lora_layers
|
47 |
+
|
48 |
+
|
49 |
+
|
50 |
+
from diffusers.utils.torch_utils import randn_tensor
|
51 |
+
from diffusers.pipelines.controlnet import StableDiffusionControlNetPipeline
|
52 |
+
from diffusers.pipelines.pipeline_utils import DiffusionPipeline
|
53 |
+
from diffusers.pipelines.marigold.marigold_image_processing import MarigoldImageProcessor
|
54 |
+
from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
|
55 |
+
import torch.nn.functional as F
|
56 |
+
|
57 |
+
import pdb
|
58 |
+
|
59 |
+
|
60 |
+
|
61 |
+
logger = logging.get_logger(__name__) # pylint: disable=invalid-name
|
62 |
+
|
63 |
+
|
64 |
+
EXAMPLE_DOC_STRING = """
|
65 |
+
Examples:
|
66 |
+
```py
|
67 |
+
>>> import diffusers
|
68 |
+
>>> import torch
|
69 |
+
|
70 |
+
>>> pipe = diffusers.MarigoldNormalsPipeline.from_pretrained(
|
71 |
+
... "prs-eth/marigold-normals-lcm-v0-1", variant="fp16", torch_dtype=torch.float16
|
72 |
+
... ).to("cuda")
|
73 |
+
|
74 |
+
>>> image = diffusers.utils.load_image("https://marigoldmonodepth.github.io/images/einstein.jpg")
|
75 |
+
>>> normals = pipe(image)
|
76 |
+
|
77 |
+
>>> vis = pipe.image_processor.visualize_normals(normals.prediction)
|
78 |
+
>>> vis[0].save("einstein_normals.png")
|
79 |
+
```
|
80 |
+
"""
|
81 |
+
|
82 |
+
|
83 |
+
@dataclass
|
84 |
+
class StableNormalOutput(BaseOutput):
|
85 |
+
"""
|
86 |
+
Output class for Marigold monocular normals prediction pipeline.
|
87 |
+
|
88 |
+
Args:
|
89 |
+
prediction (`np.ndarray`, `torch.Tensor`):
|
90 |
+
Predicted normals with values in the range [-1, 1]. The shape is always $numimages \times 3 \times height
|
91 |
+
\times width$, regardless of whether the images were passed as a 4D array or a list.
|
92 |
+
uncertainty (`None`, `np.ndarray`, `torch.Tensor`):
|
93 |
+
Uncertainty maps computed from the ensemble, with values in the range [0, 1]. The shape is $numimages
|
94 |
+
\times 1 \times height \times width$.
|
95 |
+
latent (`None`, `torch.Tensor`):
|
96 |
+
Latent features corresponding to the predictions, compatible with the `latents` argument of the pipeline.
|
97 |
+
The shape is $numimages * numensemble \times 4 \times latentheight \times latentwidth$.
|
98 |
+
"""
|
99 |
+
|
100 |
+
prediction: Union[np.ndarray, torch.Tensor]
|
101 |
+
latent: Union[None, torch.Tensor]
|
102 |
+
|
103 |
+
|
104 |
+
class StableNormalPipeline(StableDiffusionControlNetPipeline):
|
105 |
+
""" Pipeline for monocular normals estimation using the Marigold method: https://marigoldmonodepth.github.io.
|
106 |
+
Pipeline for text-to-image generation using Stable Diffusion with ControlNet guidance.
|
107 |
+
|
108 |
+
This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
|
109 |
+
implemented for all pipelines (downloading, saving, running on a particular device, etc.).
|
110 |
+
|
111 |
+
The pipeline also inherits the following loading methods:
|
112 |
+
- [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
|
113 |
+
- [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
|
114 |
+
- [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
|
115 |
+
- [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
|
116 |
+
- [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
|
117 |
+
|
118 |
+
Args:
|
119 |
+
vae ([`AutoencoderKL`]):
|
120 |
+
Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations.
|
121 |
+
text_encoder ([`~transformers.CLIPTextModel`]):
|
122 |
+
Frozen text-encoder ([clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14)).
|
123 |
+
tokenizer ([`~transformers.CLIPTokenizer`]):
|
124 |
+
A `CLIPTokenizer` to tokenize text.
|
125 |
+
unet ([`UNet2DConditionModel`]):
|
126 |
+
A `UNet2DConditionModel` to denoise the encoded image latents.
|
127 |
+
controlnet ([`ControlNetModel`] or `List[ControlNetModel]`):
|
128 |
+
Provides additional conditioning to the `unet` during the denoising process. If you set multiple
|
129 |
+
ControlNets as a list, the outputs from each ControlNet are added together to create one combined
|
130 |
+
additional conditioning.
|
131 |
+
scheduler ([`SchedulerMixin`]):
|
132 |
+
A scheduler to be used in combination with `unet` to denoise the encoded image latents. Can be one of
|
133 |
+
[`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
|
134 |
+
safety_checker ([`StableDiffusionSafetyChecker`]):
|
135 |
+
Classification module that estimates whether generated images could be considered offensive or harmful.
|
136 |
+
Please refer to the [model card](https://huggingface.co/runwayml/stable-diffusion-v1-5) for more details
|
137 |
+
about a model's potential harms.
|
138 |
+
feature_extractor ([`~transformers.CLIPImageProcessor`]):
|
139 |
+
A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
|
140 |
+
"""
|
141 |
+
|
142 |
+
model_cpu_offload_seq = "text_encoder->image_encoder->unet->vae"
|
143 |
+
_optional_components = ["safety_checker", "feature_extractor", "image_encoder"]
|
144 |
+
_exclude_from_cpu_offload = ["safety_checker"]
|
145 |
+
_callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"]
|
146 |
+
|
147 |
+
|
148 |
+
|
149 |
+
def __init__(
|
150 |
+
self,
|
151 |
+
vae: AutoencoderKL,
|
152 |
+
text_encoder: CLIPTextModel,
|
153 |
+
tokenizer: CLIPTokenizer,
|
154 |
+
unet: UNet2DConditionModel,
|
155 |
+
controlnet: Union[ControlNetModel, List[ControlNetModel], Tuple[ControlNetModel]],
|
156 |
+
dino_controlnet: Union[ControlNetModel, List[ControlNetModel], Tuple[ControlNetModel]],
|
157 |
+
scheduler: Union[DDIMScheduler],
|
158 |
+
safety_checker: StableDiffusionSafetyChecker,
|
159 |
+
feature_extractor: CLIPImageProcessor,
|
160 |
+
image_encoder: CLIPVisionModelWithProjection = None,
|
161 |
+
requires_safety_checker: bool = True,
|
162 |
+
default_denoising_steps: Optional[int] = 10,
|
163 |
+
default_processing_resolution: Optional[int] = 768,
|
164 |
+
prompt="The normal map",
|
165 |
+
empty_text_embedding=None,
|
166 |
+
t_start: Optional[int] = 401,
|
167 |
+
):
|
168 |
+
super().__init__(
|
169 |
+
vae,
|
170 |
+
text_encoder,
|
171 |
+
tokenizer,
|
172 |
+
unet,
|
173 |
+
controlnet,
|
174 |
+
scheduler,
|
175 |
+
safety_checker,
|
176 |
+
feature_extractor,
|
177 |
+
image_encoder,
|
178 |
+
requires_safety_checker,
|
179 |
+
)
|
180 |
+
|
181 |
+
self.register_modules(
|
182 |
+
dino_controlnet=dino_controlnet,
|
183 |
+
)
|
184 |
+
|
185 |
+
self.vae_scale_factor = 768
|
186 |
+
self.image_processor = MarigoldImageProcessor(vae_scale_factor=self.vae_scale_factor)
|
187 |
+
self.dino_image_processor = lambda x: x / 127.5 -1.
|
188 |
+
|
189 |
+
self.default_denoising_steps = default_denoising_steps
|
190 |
+
self.default_processing_resolution = default_processing_resolution
|
191 |
+
self.prompt = prompt
|
192 |
+
self.prompt_embeds = None
|
193 |
+
self.empty_text_embedding = empty_text_embedding
|
194 |
+
self.t_start= torch.tensor(t_start) # target_out latents
|
195 |
+
|
196 |
+
|
197 |
+
def check_inputs(
|
198 |
+
self,
|
199 |
+
image: PipelineImageInput,
|
200 |
+
num_inference_steps: int,
|
201 |
+
ensemble_size: int,
|
202 |
+
processing_resolution: int,
|
203 |
+
resample_method_input: str,
|
204 |
+
resample_method_output: str,
|
205 |
+
batch_size: int,
|
206 |
+
ensembling_kwargs: Optional[Dict[str, Any]],
|
207 |
+
latents: Optional[torch.Tensor],
|
208 |
+
generator: Optional[Union[torch.Generator, List[torch.Generator]]],
|
209 |
+
output_type: str,
|
210 |
+
output_uncertainty: bool,
|
211 |
+
) -> int:
|
212 |
+
if num_inference_steps is None:
|
213 |
+
raise ValueError("`num_inference_steps` is not specified and could not be resolved from the model config.")
|
214 |
+
if num_inference_steps < 1:
|
215 |
+
raise ValueError("`num_inference_steps` must be positive.")
|
216 |
+
if ensemble_size < 1:
|
217 |
+
raise ValueError("`ensemble_size` must be positive.")
|
218 |
+
if ensemble_size == 2:
|
219 |
+
logger.warning(
|
220 |
+
"`ensemble_size` == 2 results are similar to no ensembling (1); "
|
221 |
+
"consider increasing the value to at least 3."
|
222 |
+
)
|
223 |
+
if ensemble_size == 1 and output_uncertainty:
|
224 |
+
raise ValueError(
|
225 |
+
"Computing uncertainty by setting `output_uncertainty=True` also requires setting `ensemble_size` "
|
226 |
+
"greater than 1."
|
227 |
+
)
|
228 |
+
if processing_resolution is None:
|
229 |
+
raise ValueError(
|
230 |
+
"`processing_resolution` is not specified and could not be resolved from the model config."
|
231 |
+
)
|
232 |
+
if processing_resolution < 0:
|
233 |
+
raise ValueError(
|
234 |
+
"`processing_resolution` must be non-negative: 0 for native resolution, or any positive value for "
|
235 |
+
"downsampled processing."
|
236 |
+
)
|
237 |
+
if processing_resolution % self.vae_scale_factor != 0:
|
238 |
+
raise ValueError(f"`processing_resolution` must be a multiple of {self.vae_scale_factor}.")
|
239 |
+
if resample_method_input not in ("nearest", "nearest-exact", "bilinear", "bicubic", "area"):
|
240 |
+
raise ValueError(
|
241 |
+
"`resample_method_input` takes string values compatible with PIL library: "
|
242 |
+
"nearest, nearest-exact, bilinear, bicubic, area."
|
243 |
+
)
|
244 |
+
if resample_method_output not in ("nearest", "nearest-exact", "bilinear", "bicubic", "area"):
|
245 |
+
raise ValueError(
|
246 |
+
"`resample_method_output` takes string values compatible with PIL library: "
|
247 |
+
"nearest, nearest-exact, bilinear, bicubic, area."
|
248 |
+
)
|
249 |
+
if batch_size < 1:
|
250 |
+
raise ValueError("`batch_size` must be positive.")
|
251 |
+
if output_type not in ["pt", "np"]:
|
252 |
+
raise ValueError("`output_type` must be one of `pt` or `np`.")
|
253 |
+
if latents is not None and generator is not None:
|
254 |
+
raise ValueError("`latents` and `generator` cannot be used together.")
|
255 |
+
if ensembling_kwargs is not None:
|
256 |
+
if not isinstance(ensembling_kwargs, dict):
|
257 |
+
raise ValueError("`ensembling_kwargs` must be a dictionary.")
|
258 |
+
if "reduction" in ensembling_kwargs and ensembling_kwargs["reduction"] not in ("closest", "mean"):
|
259 |
+
raise ValueError("`ensembling_kwargs['reduction']` can be either `'closest'` or `'mean'`.")
|
260 |
+
|
261 |
+
# image checks
|
262 |
+
num_images = 0
|
263 |
+
W, H = None, None
|
264 |
+
if not isinstance(image, list):
|
265 |
+
image = [image]
|
266 |
+
for i, img in enumerate(image):
|
267 |
+
if isinstance(img, np.ndarray) or torch.is_tensor(img):
|
268 |
+
if img.ndim not in (2, 3, 4):
|
269 |
+
raise ValueError(f"`image[{i}]` has unsupported dimensions or shape: {img.shape}.")
|
270 |
+
H_i, W_i = img.shape[-2:]
|
271 |
+
N_i = 1
|
272 |
+
if img.ndim == 4:
|
273 |
+
N_i = img.shape[0]
|
274 |
+
elif isinstance(img, Image.Image):
|
275 |
+
W_i, H_i = img.size
|
276 |
+
N_i = 1
|
277 |
+
else:
|
278 |
+
raise ValueError(f"Unsupported `image[{i}]` type: {type(img)}.")
|
279 |
+
if W is None:
|
280 |
+
W, H = W_i, H_i
|
281 |
+
elif (W, H) != (W_i, H_i):
|
282 |
+
raise ValueError(
|
283 |
+
f"Input `image[{i}]` has incompatible dimensions {(W_i, H_i)} with the previous images {(W, H)}"
|
284 |
+
)
|
285 |
+
num_images += N_i
|
286 |
+
|
287 |
+
# latents checks
|
288 |
+
if latents is not None:
|
289 |
+
if not torch.is_tensor(latents):
|
290 |
+
raise ValueError("`latents` must be a torch.Tensor.")
|
291 |
+
if latents.dim() != 4:
|
292 |
+
raise ValueError(f"`latents` has unsupported dimensions or shape: {latents.shape}.")
|
293 |
+
|
294 |
+
if processing_resolution > 0:
|
295 |
+
max_orig = max(H, W)
|
296 |
+
new_H = H * processing_resolution // max_orig
|
297 |
+
new_W = W * processing_resolution // max_orig
|
298 |
+
if new_H == 0 or new_W == 0:
|
299 |
+
raise ValueError(f"Extreme aspect ratio of the input image: [{W} x {H}]")
|
300 |
+
W, H = new_W, new_H
|
301 |
+
w = (W + self.vae_scale_factor - 1) // self.vae_scale_factor
|
302 |
+
h = (H + self.vae_scale_factor - 1) // self.vae_scale_factor
|
303 |
+
shape_expected = (num_images * ensemble_size, self.vae.config.latent_channels, h, w)
|
304 |
+
|
305 |
+
if latents.shape != shape_expected:
|
306 |
+
raise ValueError(f"`latents` has unexpected shape={latents.shape} expected={shape_expected}.")
|
307 |
+
|
308 |
+
# generator checks
|
309 |
+
if generator is not None:
|
310 |
+
if isinstance(generator, list):
|
311 |
+
if len(generator) != num_images * ensemble_size:
|
312 |
+
raise ValueError(
|
313 |
+
"The number of generators must match the total number of ensemble members for all input images."
|
314 |
+
)
|
315 |
+
if not all(g.device.type == generator[0].device.type for g in generator):
|
316 |
+
raise ValueError("`generator` device placement is not consistent in the list.")
|
317 |
+
elif not isinstance(generator, torch.Generator):
|
318 |
+
raise ValueError(f"Unsupported generator type: {type(generator)}.")
|
319 |
+
|
320 |
+
return num_images
|
321 |
+
|
322 |
+
def progress_bar(self, iterable=None, total=None, desc=None, leave=True):
|
323 |
+
if not hasattr(self, "_progress_bar_config"):
|
324 |
+
self._progress_bar_config = {}
|
325 |
+
elif not isinstance(self._progress_bar_config, dict):
|
326 |
+
raise ValueError(
|
327 |
+
f"`self._progress_bar_config` should be of type `dict`, but is {type(self._progress_bar_config)}."
|
328 |
+
)
|
329 |
+
|
330 |
+
progress_bar_config = dict(**self._progress_bar_config)
|
331 |
+
progress_bar_config["desc"] = progress_bar_config.get("desc", desc)
|
332 |
+
progress_bar_config["leave"] = progress_bar_config.get("leave", leave)
|
333 |
+
if iterable is not None:
|
334 |
+
return tqdm(iterable, **progress_bar_config)
|
335 |
+
elif total is not None:
|
336 |
+
return tqdm(total=total, **progress_bar_config)
|
337 |
+
else:
|
338 |
+
raise ValueError("Either `total` or `iterable` has to be defined.")
|
339 |
+
|
340 |
+
@torch.no_grad()
|
341 |
+
@replace_example_docstring(EXAMPLE_DOC_STRING)
|
342 |
+
def __call__(
|
343 |
+
self,
|
344 |
+
image: PipelineImageInput,
|
345 |
+
prompt: Union[str, List[str]] = None,
|
346 |
+
negative_prompt: Optional[Union[str, List[str]]] = None,
|
347 |
+
num_inference_steps: Optional[int] = None,
|
348 |
+
ensemble_size: int = 1,
|
349 |
+
processing_resolution: Optional[int] = None,
|
350 |
+
return_intermediate_result: bool = False,
|
351 |
+
match_input_resolution: bool = True,
|
352 |
+
resample_method_input: str = "bilinear",
|
353 |
+
resample_method_output: str = "bilinear",
|
354 |
+
batch_size: int = 1,
|
355 |
+
ensembling_kwargs: Optional[Dict[str, Any]] = None,
|
356 |
+
latents: Optional[Union[torch.Tensor, List[torch.Tensor]]] = None,
|
357 |
+
prompt_embeds: Optional[torch.Tensor] = None,
|
358 |
+
negative_prompt_embeds: Optional[torch.Tensor] = None,
|
359 |
+
num_images_per_prompt: Optional[int] = 1,
|
360 |
+
generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
|
361 |
+
controlnet_conditioning_scale: Union[float, List[float]] = 1.0,
|
362 |
+
output_type: str = "np",
|
363 |
+
output_uncertainty: bool = False,
|
364 |
+
output_latent: bool = False,
|
365 |
+
return_dict: bool = True,
|
366 |
+
):
|
367 |
+
"""
|
368 |
+
Function invoked when calling the pipeline.
|
369 |
+
|
370 |
+
Args:
|
371 |
+
image (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`),
|
372 |
+
`List[torch.Tensor]`: An input image or images used as an input for the normals estimation task. For
|
373 |
+
arrays and tensors, the expected value range is between `[0, 1]`. Passing a batch of images is possible
|
374 |
+
by providing a four-dimensional array or a tensor. Additionally, a list of images of two- or
|
375 |
+
three-dimensional arrays or tensors can be passed. In the latter case, all list elements must have the
|
376 |
+
same width and height.
|
377 |
+
num_inference_steps (`int`, *optional*, defaults to `None`):
|
378 |
+
Number of denoising diffusion steps during inference. The default value `None` results in automatic
|
379 |
+
selection. The number of steps should be at least 10 with the full Marigold models, and between 1 and 4
|
380 |
+
for Marigold-LCM models.
|
381 |
+
ensemble_size (`int`, defaults to `1`):
|
382 |
+
Number of ensemble predictions. Recommended values are 5 and higher for better precision, or 1 for
|
383 |
+
faster inference.
|
384 |
+
processing_resolution (`int`, *optional*, defaults to `None`):
|
385 |
+
Effective processing resolution. When set to `0`, matches the larger input image dimension. This
|
386 |
+
produces crisper predictions, but may also lead to the overall loss of global context. The default
|
387 |
+
value `None` resolves to the optimal value from the model config.
|
388 |
+
match_input_resolution (`bool`, *optional*, defaults to `True`):
|
389 |
+
When enabled, the output prediction is resized to match the input dimensions. When disabled, the longer
|
390 |
+
side of the output will equal to `processing_resolution`.
|
391 |
+
resample_method_input (`str`, *optional*, defaults to `"bilinear"`):
|
392 |
+
Resampling method used to resize input images to `processing_resolution`. The accepted values are:
|
393 |
+
`"nearest"`, `"nearest-exact"`, `"bilinear"`, `"bicubic"`, or `"area"`.
|
394 |
+
resample_method_output (`str`, *optional*, defaults to `"bilinear"`):
|
395 |
+
Resampling method used to resize output predictions to match the input resolution. The accepted values
|
396 |
+
are `"nearest"`, `"nearest-exact"`, `"bilinear"`, `"bicubic"`, or `"area"`.
|
397 |
+
batch_size (`int`, *optional*, defaults to `1`):
|
398 |
+
Batch size; only matters when setting `ensemble_size` or passing a tensor of images.
|
399 |
+
ensembling_kwargs (`dict`, *optional*, defaults to `None`)
|
400 |
+
Extra dictionary with arguments for precise ensembling control. The following options are available:
|
401 |
+
- reduction (`str`, *optional*, defaults to `"closest"`): Defines the ensembling function applied in
|
402 |
+
every pixel location, can be either `"closest"` or `"mean"`.
|
403 |
+
latents (`torch.Tensor`, *optional*, defaults to `None`):
|
404 |
+
Latent noise tensors to replace the random initialization. These can be taken from the previous
|
405 |
+
function call's output.
|
406 |
+
generator (`torch.Generator`, or `List[torch.Generator]`, *optional*, defaults to `None`):
|
407 |
+
Random number generator object to ensure reproducibility.
|
408 |
+
output_type (`str`, *optional*, defaults to `"np"`):
|
409 |
+
Preferred format of the output's `prediction` and the optional `uncertainty` fields. The accepted
|
410 |
+
values are: `"np"` (numpy array) or `"pt"` (torch tensor).
|
411 |
+
output_uncertainty (`bool`, *optional*, defaults to `False`):
|
412 |
+
When enabled, the output's `uncertainty` field contains the predictive uncertainty map, provided that
|
413 |
+
the `ensemble_size` argument is set to a value above 2.
|
414 |
+
output_latent (`bool`, *optional*, defaults to `False`):
|
415 |
+
When enabled, the output's `latent` field contains the latent codes corresponding to the predictions
|
416 |
+
within the ensemble. These codes can be saved, modified, and used for subsequent calls with the
|
417 |
+
`latents` argument.
|
418 |
+
return_dict (`bool`, *optional*, defaults to `True`):
|
419 |
+
Whether or not to return a [`~pipelines.marigold.MarigoldDepthOutput`] instead of a plain tuple.
|
420 |
+
|
421 |
+
Examples:
|
422 |
+
|
423 |
+
Returns:
|
424 |
+
[`~pipelines.marigold.MarigoldNormalsOutput`] or `tuple`:
|
425 |
+
If `return_dict` is `True`, [`~pipelines.marigold.MarigoldNormalsOutput`] is returned, otherwise a
|
426 |
+
`tuple` is returned where the first element is the prediction, the second element is the uncertainty
|
427 |
+
(or `None`), and the third is the latent (or `None`).
|
428 |
+
"""
|
429 |
+
|
430 |
+
# 0. Resolving variables.
|
431 |
+
device = self._execution_device
|
432 |
+
dtype = self.dtype
|
433 |
+
|
434 |
+
# Model-specific optimal default values leading to fast and reasonable results.
|
435 |
+
if num_inference_steps is None:
|
436 |
+
num_inference_steps = self.default_denoising_steps
|
437 |
+
if processing_resolution is None:
|
438 |
+
processing_resolution = self.default_processing_resolution
|
439 |
+
|
440 |
+
|
441 |
+
image, padding, original_resolution = self.image_processor.preprocess(
|
442 |
+
image, processing_resolution, resample_method_input, device, dtype
|
443 |
+
) # [N,3,PPH,PPW]
|
444 |
+
|
445 |
+
# 0. X_start latent obtain
|
446 |
+
predictor = self.x_start_pipeline(image, skip_preprocess=True)
|
447 |
+
x_start_latent = predictor.latent
|
448 |
+
gauss_latent = predictor.gauss_latent
|
449 |
+
|
450 |
+
# 1. Check inputs.
|
451 |
+
num_images = self.check_inputs(
|
452 |
+
image,
|
453 |
+
num_inference_steps,
|
454 |
+
ensemble_size,
|
455 |
+
processing_resolution,
|
456 |
+
resample_method_input,
|
457 |
+
resample_method_output,
|
458 |
+
batch_size,
|
459 |
+
ensembling_kwargs,
|
460 |
+
latents,
|
461 |
+
generator,
|
462 |
+
output_type,
|
463 |
+
output_uncertainty,
|
464 |
+
)
|
465 |
+
|
466 |
+
|
467 |
+
# 2. Prepare empty text conditioning.
|
468 |
+
# Model invocation: self.tokenizer, self.text_encoder.
|
469 |
+
if self.empty_text_embedding is None:
|
470 |
+
prompt = ""
|
471 |
+
text_inputs = self.tokenizer(
|
472 |
+
prompt,
|
473 |
+
padding="do_not_pad",
|
474 |
+
max_length=self.tokenizer.model_max_length,
|
475 |
+
truncation=True,
|
476 |
+
return_tensors="pt",
|
477 |
+
)
|
478 |
+
text_input_ids = text_inputs.input_ids.to(device)
|
479 |
+
self.empty_text_embedding = self.text_encoder(text_input_ids)[0] # [1,2,1024]
|
480 |
+
|
481 |
+
|
482 |
+
|
483 |
+
# 3. prepare prompt
|
484 |
+
if self.prompt_embeds is None:
|
485 |
+
prompt_embeds, negative_prompt_embeds = self.encode_prompt(
|
486 |
+
self.prompt,
|
487 |
+
device,
|
488 |
+
num_images_per_prompt,
|
489 |
+
False,
|
490 |
+
negative_prompt,
|
491 |
+
prompt_embeds=prompt_embeds,
|
492 |
+
negative_prompt_embeds=None,
|
493 |
+
lora_scale=None,
|
494 |
+
clip_skip=None,
|
495 |
+
)
|
496 |
+
self.prompt_embeds = prompt_embeds
|
497 |
+
self.negative_prompt_embeds = negative_prompt_embeds
|
498 |
+
|
499 |
+
|
500 |
+
|
501 |
+
# 5. dino guider features obtaining
|
502 |
+
## TODO different case-1
|
503 |
+
dino_features = self.prior(image)
|
504 |
+
dino_features = self.dino_controlnet.dino_controlnet_cond_embedding(dino_features)
|
505 |
+
dino_features = self.match_noisy(dino_features, x_start_latent)
|
506 |
+
|
507 |
+
# 6. Encode input image into latent space. At this step, each of the `N` input images is represented with `E`
|
508 |
+
# ensemble members. Each ensemble member is an independent diffused prediction, just initialized independently.
|
509 |
+
# Latents of each such predictions across all input images and all ensemble members are represented in the
|
510 |
+
# `pred_latent` variable. The variable `image_latent` is of the same shape: it contains each input image encoded
|
511 |
+
# into latent space and replicated `E` times. The latents can be either generated (see `generator` to ensure
|
512 |
+
# reproducibility), or passed explicitly via the `latents` argument. The latter can be set outside the pipeline
|
513 |
+
# code. For example, in the Marigold-LCM video processing demo, the latents initialization of a frame is taken
|
514 |
+
# as a convex combination of the latents output of the pipeline for the previous frame and a newly-sampled
|
515 |
+
# noise. This behavior can be achieved by setting the `output_latent` argument to `True`. The latent space
|
516 |
+
# dimensions are `(h, w)`. Encoding into latent space happens in batches of size `batch_size`.
|
517 |
+
# Model invocation: self.vae.encoder.
|
518 |
+
image_latent, pred_latent = self.prepare_latents(
|
519 |
+
image, latents, generator, ensemble_size, batch_size
|
520 |
+
) # [N*E,4,h,w], [N*E,4,h,w]
|
521 |
+
|
522 |
+
|
523 |
+
del (
|
524 |
+
image,
|
525 |
+
)
|
526 |
+
|
527 |
+
# 7. denoise sampling, using heuritic sampling proposed by Ye.
|
528 |
+
self.scheduler.set_timesteps(num_inference_steps, device=device)
|
529 |
+
|
530 |
+
cond_scale =controlnet_conditioning_scale
|
531 |
+
pred_latent = x_start_latent
|
532 |
+
|
533 |
+
cur_step = 0
|
534 |
+
|
535 |
+
# dino controlnet
|
536 |
+
dino_down_block_res_samples, dino_mid_block_res_sample = self.dino_controlnet(
|
537 |
+
dino_features.detach(),
|
538 |
+
0, # not depend on time steps
|
539 |
+
encoder_hidden_states=self.prompt_embeds,
|
540 |
+
conditioning_scale=cond_scale,
|
541 |
+
guess_mode=False,
|
542 |
+
return_dict=False,
|
543 |
+
)
|
544 |
+
assert dino_mid_block_res_sample == None
|
545 |
+
|
546 |
+
pred_latents = []
|
547 |
+
|
548 |
+
down_block_res_samples, mid_block_res_sample = self.controlnet(
|
549 |
+
image_latent.detach(),
|
550 |
+
self.t_start,
|
551 |
+
encoder_hidden_states=self.prompt_embeds,
|
552 |
+
conditioning_scale=cond_scale,
|
553 |
+
guess_mode=False,
|
554 |
+
return_dict=False,
|
555 |
+
)
|
556 |
+
last_pred_latent = pred_latent
|
557 |
+
for i in range(4):
|
558 |
+
_dino_down_block_res_samples = [dino_down_block_res_sample for dino_down_block_res_sample in dino_down_block_res_samples] # copy, avoid repeat quiery
|
559 |
+
|
560 |
+
model_output = self.dino_unet_forward(
|
561 |
+
self.unet,
|
562 |
+
pred_latent,
|
563 |
+
self.t_start,
|
564 |
+
encoder_hidden_states=self.prompt_embeds,
|
565 |
+
down_block_additional_residuals=down_block_res_samples,
|
566 |
+
mid_block_additional_residual=mid_block_res_sample,
|
567 |
+
dino_down_block_additional_residuals= _dino_down_block_res_samples,
|
568 |
+
return_dict=False,
|
569 |
+
)[0] # [B,4,h,w]
|
570 |
+
pred_latents.append(model_output)
|
571 |
+
pred_latent = self.scheduler.add_noise(model_output, gauss_latent, self.t_start)
|
572 |
+
pred_latent = 0.4 * pred_latent + 0.6 * last_pred_latent
|
573 |
+
last_pred_latent = pred_latent
|
574 |
+
pred_latents = torch.cat(pred_latents, dim=0)
|
575 |
+
del (
|
576 |
+
image_latent,
|
577 |
+
dino_features,
|
578 |
+
)
|
579 |
+
|
580 |
+
|
581 |
+
# decoder
|
582 |
+
if return_intermediate_result:
|
583 |
+
prediction = []
|
584 |
+
for _pred_latent in pred_latents:
|
585 |
+
_prediction = self.decode_prediction(_pred_latent.unsqueeze(dim=0))
|
586 |
+
prediction.append(_prediction)
|
587 |
+
prediction = torch.cat(prediction, dim=0)
|
588 |
+
else:
|
589 |
+
prediction = self.decode_prediction(pred_latents[-1].unsqueeze(dim=0))
|
590 |
+
prediction = self.image_processor.unpad_image(prediction, padding) # [N*E,3,PH,PW]
|
591 |
+
|
592 |
+
if match_input_resolution:
|
593 |
+
prediction = self.image_processor.resize_antialias(
|
594 |
+
prediction, original_resolution, resample_method_output, is_aa=False
|
595 |
+
) # [N,3,H,W]
|
596 |
+
prediction = self.normalize_normals(prediction) # [N,3,H,W]
|
597 |
+
|
598 |
+
if output_type == "np":
|
599 |
+
prediction = self.image_processor.pt_to_numpy(prediction) # [N,H,W,3]
|
600 |
+
prediction = prediction.clip(min=-1, max=1)
|
601 |
+
|
602 |
+
# 11. Offload all models
|
603 |
+
self.maybe_free_model_hooks()
|
604 |
+
|
605 |
+
return StableNormalOutput(
|
606 |
+
prediction=prediction,
|
607 |
+
latent=pred_latent,
|
608 |
+
)
|
609 |
+
|
610 |
+
# Copied from diffusers.pipelines.marigold.pipeline_marigold_depth.MarigoldDepthPipeline.prepare_latents
|
611 |
+
def prepare_latents(
|
612 |
+
self,
|
613 |
+
image: torch.Tensor,
|
614 |
+
latents: Optional[torch.Tensor],
|
615 |
+
generator: Optional[torch.Generator],
|
616 |
+
ensemble_size: int,
|
617 |
+
batch_size: int,
|
618 |
+
) -> Tuple[torch.Tensor, torch.Tensor]:
|
619 |
+
def retrieve_latents(encoder_output):
|
620 |
+
if hasattr(encoder_output, "latent_dist"):
|
621 |
+
return encoder_output.latent_dist.mode()
|
622 |
+
elif hasattr(encoder_output, "latents"):
|
623 |
+
return encoder_output.latents
|
624 |
+
else:
|
625 |
+
raise AttributeError("Could not access latents of provided encoder_output")
|
626 |
+
|
627 |
+
|
628 |
+
|
629 |
+
image_latent = torch.cat(
|
630 |
+
[
|
631 |
+
retrieve_latents(self.vae.encode(image[i : i + batch_size]))
|
632 |
+
for i in range(0, image.shape[0], batch_size)
|
633 |
+
],
|
634 |
+
dim=0,
|
635 |
+
) # [N,4,h,w]
|
636 |
+
image_latent = image_latent * self.vae.config.scaling_factor
|
637 |
+
image_latent = image_latent.repeat_interleave(ensemble_size, dim=0) # [N*E,4,h,w]
|
638 |
+
|
639 |
+
pred_latent = latents
|
640 |
+
if pred_latent is None:
|
641 |
+
|
642 |
+
|
643 |
+
pred_latent = randn_tensor(
|
644 |
+
image_latent.shape,
|
645 |
+
generator=generator,
|
646 |
+
device=image_latent.device,
|
647 |
+
dtype=image_latent.dtype,
|
648 |
+
) # [N*E,4,h,w]
|
649 |
+
|
650 |
+
return image_latent, pred_latent
|
651 |
+
|
652 |
+
def decode_prediction(self, pred_latent: torch.Tensor) -> torch.Tensor:
|
653 |
+
if pred_latent.dim() != 4 or pred_latent.shape[1] != self.vae.config.latent_channels:
|
654 |
+
raise ValueError(
|
655 |
+
f"Expecting 4D tensor of shape [B,{self.vae.config.latent_channels},H,W]; got {pred_latent.shape}."
|
656 |
+
)
|
657 |
+
|
658 |
+
prediction = self.vae.decode(pred_latent / self.vae.config.scaling_factor, return_dict=False)[0] # [B,3,H,W]
|
659 |
+
|
660 |
+
return prediction # [B,3,H,W]
|
661 |
+
|
662 |
+
@staticmethod
|
663 |
+
def normalize_normals(normals: torch.Tensor, eps: float = 1e-6) -> torch.Tensor:
|
664 |
+
if normals.dim() != 4 or normals.shape[1] != 3:
|
665 |
+
raise ValueError(f"Expecting 4D tensor of shape [B,3,H,W]; got {normals.shape}.")
|
666 |
+
|
667 |
+
norm = torch.norm(normals, dim=1, keepdim=True)
|
668 |
+
normals /= norm.clamp(min=eps)
|
669 |
+
|
670 |
+
return normals
|
671 |
+
|
672 |
+
@staticmethod
|
673 |
+
def match_noisy(dino, noisy):
|
674 |
+
_, __, dino_h, dino_w = dino.shape
|
675 |
+
_, __, h, w = noisy.shape
|
676 |
+
|
677 |
+
if h == dino_h and w == dino_w:
|
678 |
+
return dino
|
679 |
+
else:
|
680 |
+
return F.interpolate(dino, (h, w), mode='bilinear')
|
681 |
+
|
682 |
+
|
683 |
+
|
684 |
+
|
685 |
+
|
686 |
+
|
687 |
+
|
688 |
+
|
689 |
+
|
690 |
+
|
691 |
+
@staticmethod
|
692 |
+
def dino_unet_forward(
|
693 |
+
self, # NOTE that repurpose to UNet
|
694 |
+
sample: torch.Tensor,
|
695 |
+
timestep: Union[torch.Tensor, float, int],
|
696 |
+
encoder_hidden_states: torch.Tensor,
|
697 |
+
class_labels: Optional[torch.Tensor] = None,
|
698 |
+
timestep_cond: Optional[torch.Tensor] = None,
|
699 |
+
attention_mask: Optional[torch.Tensor] = None,
|
700 |
+
cross_attention_kwargs: Optional[Dict[str, Any]] = None,
|
701 |
+
added_cond_kwargs: Optional[Dict[str, torch.Tensor]] = None,
|
702 |
+
down_block_additional_residuals: Optional[Tuple[torch.Tensor]] = None,
|
703 |
+
mid_block_additional_residual: Optional[torch.Tensor] = None,
|
704 |
+
dino_down_block_additional_residuals: Optional[torch.Tensor] = None,
|
705 |
+
down_intrablock_additional_residuals: Optional[Tuple[torch.Tensor]] = None,
|
706 |
+
encoder_attention_mask: Optional[torch.Tensor] = None,
|
707 |
+
return_dict: bool = True,
|
708 |
+
) -> Union[UNet2DConditionOutput, Tuple]:
|
709 |
+
r"""
|
710 |
+
The [`UNet2DConditionModel`] forward method.
|
711 |
+
|
712 |
+
Args:
|
713 |
+
sample (`torch.Tensor`):
|
714 |
+
The noisy input tensor with the following shape `(batch, channel, height, width)`.
|
715 |
+
timestep (`torch.Tensor` or `float` or `int`): The number of timesteps to denoise an input.
|
716 |
+
encoder_hidden_states (`torch.Tensor`):
|
717 |
+
The encoder hidden states with shape `(batch, sequence_length, feature_dim)`.
|
718 |
+
class_labels (`torch.Tensor`, *optional*, defaults to `None`):
|
719 |
+
Optional class labels for conditioning. Their embeddings will be summed with the timestep embeddings.
|
720 |
+
timestep_cond: (`torch.Tensor`, *optional*, defaults to `None`):
|
721 |
+
Conditional embeddings for timestep. If provided, the embeddings will be summed with the samples passed
|
722 |
+
through the `self.time_embedding` layer to obtain the timestep embeddings.
|
723 |
+
attention_mask (`torch.Tensor`, *optional*, defaults to `None`):
|
724 |
+
An attention mask of shape `(batch, key_tokens)` is applied to `encoder_hidden_states`. If `1` the mask
|
725 |
+
is kept, otherwise if `0` it is discarded. Mask will be converted into a bias, which adds large
|
726 |
+
negative values to the attention scores corresponding to "discard" tokens.
|
727 |
+
cross_attention_kwargs (`dict`, *optional*):
|
728 |
+
A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
|
729 |
+
`self.processor` in
|
730 |
+
[diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
|
731 |
+
added_cond_kwargs: (`dict`, *optional*):
|
732 |
+
A kwargs dictionary containing additional embeddings that if specified are added to the embeddings that
|
733 |
+
are passed along to the UNet blocks.
|
734 |
+
down_block_additional_residuals: (`tuple` of `torch.Tensor`, *optional*):
|
735 |
+
A tuple of tensors that if specified are added to the residuals of down unet blocks.
|
736 |
+
mid_block_additional_residual: (`torch.Tensor`, *optional*):
|
737 |
+
A tensor that if specified is added to the residual of the middle unet block.
|
738 |
+
down_intrablock_additional_residuals (`tuple` of `torch.Tensor`, *optional*):
|
739 |
+
additional residuals to be added within UNet down blocks, for example from T2I-Adapter side model(s)
|
740 |
+
encoder_attention_mask (`torch.Tensor`):
|
741 |
+
A cross-attention mask of shape `(batch, sequence_length)` is applied to `encoder_hidden_states`. If
|
742 |
+
`True` the mask is kept, otherwise if `False` it is discarded. Mask will be converted into a bias,
|
743 |
+
which adds large negative values to the attention scores corresponding to "discard" tokens.
|
744 |
+
return_dict (`bool`, *optional*, defaults to `True`):
|
745 |
+
Whether or not to return a [`~models.unets.unet_2d_condition.UNet2DConditionOutput`] instead of a plain
|
746 |
+
tuple.
|
747 |
+
|
748 |
+
Returns:
|
749 |
+
[`~models.unets.unet_2d_condition.UNet2DConditionOutput`] or `tuple`:
|
750 |
+
If `return_dict` is True, an [`~models.unets.unet_2d_condition.UNet2DConditionOutput`] is returned,
|
751 |
+
otherwise a `tuple` is returned where the first element is the sample tensor.
|
752 |
+
"""
|
753 |
+
# By default samples have to be AT least a multiple of the overall upsampling factor.
|
754 |
+
# The overall upsampling factor is equal to 2 ** (# num of upsampling layers).
|
755 |
+
# However, the upsampling interpolation output size can be forced to fit any upsampling size
|
756 |
+
# on the fly if necessary.
|
757 |
+
|
758 |
+
|
759 |
+
default_overall_up_factor = 2**self.num_upsamplers
|
760 |
+
|
761 |
+
# upsample size should be forwarded when sample is not a multiple of `default_overall_up_factor`
|
762 |
+
forward_upsample_size = False
|
763 |
+
upsample_size = None
|
764 |
+
|
765 |
+
for dim in sample.shape[-2:]:
|
766 |
+
if dim % default_overall_up_factor != 0:
|
767 |
+
# Forward upsample size to force interpolation output size.
|
768 |
+
forward_upsample_size = True
|
769 |
+
break
|
770 |
+
|
771 |
+
# ensure attention_mask is a bias, and give it a singleton query_tokens dimension
|
772 |
+
# expects mask of shape:
|
773 |
+
# [batch, key_tokens]
|
774 |
+
# adds singleton query_tokens dimension:
|
775 |
+
# [batch, 1, key_tokens]
|
776 |
+
# this helps to broadcast it as a bias over attention scores, which will be in one of the following shapes:
|
777 |
+
# [batch, heads, query_tokens, key_tokens] (e.g. torch sdp attn)
|
778 |
+
# [batch * heads, query_tokens, key_tokens] (e.g. xformers or classic attn)
|
779 |
+
if attention_mask is not None:
|
780 |
+
# assume that mask is expressed as:
|
781 |
+
# (1 = keep, 0 = discard)
|
782 |
+
# convert mask into a bias that can be added to attention scores:
|
783 |
+
# (keep = +0, discard = -10000.0)
|
784 |
+
attention_mask = (1 - attention_mask.to(sample.dtype)) * -10000.0
|
785 |
+
attention_mask = attention_mask.unsqueeze(1)
|
786 |
+
|
787 |
+
# convert encoder_attention_mask to a bias the same way we do for attention_mask
|
788 |
+
if encoder_attention_mask is not None:
|
789 |
+
encoder_attention_mask = (1 - encoder_attention_mask.to(sample.dtype)) * -10000.0
|
790 |
+
encoder_attention_mask = encoder_attention_mask.unsqueeze(1)
|
791 |
+
|
792 |
+
# 0. center input if necessary
|
793 |
+
if self.config.center_input_sample:
|
794 |
+
sample = 2 * sample - 1.0
|
795 |
+
|
796 |
+
# 1. time
|
797 |
+
t_emb = self.get_time_embed(sample=sample, timestep=timestep)
|
798 |
+
emb = self.time_embedding(t_emb, timestep_cond)
|
799 |
+
aug_emb = None
|
800 |
+
|
801 |
+
class_emb = self.get_class_embed(sample=sample, class_labels=class_labels)
|
802 |
+
if class_emb is not None:
|
803 |
+
if self.config.class_embeddings_concat:
|
804 |
+
emb = torch.cat([emb, class_emb], dim=-1)
|
805 |
+
else:
|
806 |
+
emb = emb + class_emb
|
807 |
+
|
808 |
+
aug_emb = self.get_aug_embed(
|
809 |
+
emb=emb, encoder_hidden_states=encoder_hidden_states, added_cond_kwargs=added_cond_kwargs
|
810 |
+
)
|
811 |
+
if self.config.addition_embed_type == "image_hint":
|
812 |
+
aug_emb, hint = aug_emb
|
813 |
+
sample = torch.cat([sample, hint], dim=1)
|
814 |
+
|
815 |
+
emb = emb + aug_emb if aug_emb is not None else emb
|
816 |
+
|
817 |
+
if self.time_embed_act is not None:
|
818 |
+
emb = self.time_embed_act(emb)
|
819 |
+
|
820 |
+
encoder_hidden_states = self.process_encoder_hidden_states(
|
821 |
+
encoder_hidden_states=encoder_hidden_states, added_cond_kwargs=added_cond_kwargs
|
822 |
+
)
|
823 |
+
|
824 |
+
# 2. pre-process
|
825 |
+
sample = self.conv_in(sample)
|
826 |
+
|
827 |
+
# 2.5 GLIGEN position net
|
828 |
+
if cross_attention_kwargs is not None and cross_attention_kwargs.get("gligen", None) is not None:
|
829 |
+
cross_attention_kwargs = cross_attention_kwargs.copy()
|
830 |
+
gligen_args = cross_attention_kwargs.pop("gligen")
|
831 |
+
cross_attention_kwargs["gligen"] = {"objs": self.position_net(**gligen_args)}
|
832 |
+
|
833 |
+
# 3. down
|
834 |
+
# we're popping the `scale` instead of getting it because otherwise `scale` will be propagated
|
835 |
+
# to the internal blocks and will raise deprecation warnings. this will be confusing for our users.
|
836 |
+
if cross_attention_kwargs is not None:
|
837 |
+
cross_attention_kwargs = cross_attention_kwargs.copy()
|
838 |
+
lora_scale = cross_attention_kwargs.pop("scale", 1.0)
|
839 |
+
else:
|
840 |
+
lora_scale = 1.0
|
841 |
+
|
842 |
+
if USE_PEFT_BACKEND:
|
843 |
+
# weight the lora layers by setting `lora_scale` for each PEFT layer
|
844 |
+
scale_lora_layers(self, lora_scale)
|
845 |
+
|
846 |
+
is_controlnet = mid_block_additional_residual is not None and down_block_additional_residuals is not None
|
847 |
+
# using new arg down_intrablock_additional_residuals for T2I-Adapters, to distinguish from controlnets
|
848 |
+
is_adapter = down_intrablock_additional_residuals is not None
|
849 |
+
# maintain backward compatibility for legacy usage, where
|
850 |
+
# T2I-Adapter and ControlNet both use down_block_additional_residuals arg
|
851 |
+
# but can only use one or the other
|
852 |
+
if not is_adapter and mid_block_additional_residual is None and down_block_additional_residuals is not None:
|
853 |
+
deprecate(
|
854 |
+
"T2I should not use down_block_additional_residuals",
|
855 |
+
"1.3.0",
|
856 |
+
"Passing intrablock residual connections with `down_block_additional_residuals` is deprecated \
|
857 |
+
and will be removed in diffusers 1.3.0. `down_block_additional_residuals` should only be used \
|
858 |
+
for ControlNet. Please make sure use `down_intrablock_additional_residuals` instead. ",
|
859 |
+
standard_warn=False,
|
860 |
+
)
|
861 |
+
down_intrablock_additional_residuals = down_block_additional_residuals
|
862 |
+
is_adapter = True
|
863 |
+
|
864 |
+
|
865 |
+
|
866 |
+
def residual_downforward(
|
867 |
+
self, hidden_states: torch.Tensor, temb: Optional[torch.Tensor] = None,
|
868 |
+
additional_residuals: Optional[torch.Tensor] = None,
|
869 |
+
*args, **kwargs,
|
870 |
+
) -> Tuple[torch.Tensor, Tuple[torch.Tensor, ...]]:
|
871 |
+
if len(args) > 0 or kwargs.get("scale", None) is not None:
|
872 |
+
deprecation_message = "The `scale` argument is deprecated and will be ignored. Please remove it, as passing it will raise an error in the future. `scale` should directly be passed while calling the underlying pipeline component i.e., via `cross_attention_kwargs`."
|
873 |
+
deprecate("scale", "1.0.0", deprecation_message)
|
874 |
+
|
875 |
+
output_states = ()
|
876 |
+
|
877 |
+
for resnet in self.resnets:
|
878 |
+
if self.training and self.gradient_checkpointing:
|
879 |
+
|
880 |
+
def create_custom_forward(module):
|
881 |
+
def custom_forward(*inputs):
|
882 |
+
return module(*inputs)
|
883 |
+
|
884 |
+
return custom_forward
|
885 |
+
|
886 |
+
if is_torch_version(">=", "1.11.0"):
|
887 |
+
hidden_states = torch.utils.checkpoint.checkpoint(
|
888 |
+
create_custom_forward(resnet), hidden_states, temb, use_reentrant=False
|
889 |
+
)
|
890 |
+
else:
|
891 |
+
hidden_states = torch.utils.checkpoint.checkpoint(
|
892 |
+
create_custom_forward(resnet), hidden_states, temb
|
893 |
+
)
|
894 |
+
else:
|
895 |
+
hidden_states = resnet(hidden_states, temb)
|
896 |
+
hidden_states += additional_residuals.pop(0)
|
897 |
+
|
898 |
+
|
899 |
+
output_states = output_states + (hidden_states,)
|
900 |
+
|
901 |
+
if self.downsamplers is not None:
|
902 |
+
for downsampler in self.downsamplers:
|
903 |
+
hidden_states = downsampler(hidden_states)
|
904 |
+
hidden_states += additional_residuals.pop(0)
|
905 |
+
|
906 |
+
output_states = output_states + (hidden_states,)
|
907 |
+
|
908 |
+
return hidden_states, output_states
|
909 |
+
|
910 |
+
|
911 |
+
def residual_blockforward(
|
912 |
+
self, ## NOTE that repurpose to unet_blocks
|
913 |
+
hidden_states: torch.Tensor,
|
914 |
+
temb: Optional[torch.Tensor] = None,
|
915 |
+
encoder_hidden_states: Optional[torch.Tensor] = None,
|
916 |
+
attention_mask: Optional[torch.Tensor] = None,
|
917 |
+
cross_attention_kwargs: Optional[Dict[str, Any]] = None,
|
918 |
+
encoder_attention_mask: Optional[torch.Tensor] = None,
|
919 |
+
additional_residuals: Optional[torch.Tensor] = None,
|
920 |
+
) -> Tuple[torch.Tensor, Tuple[torch.Tensor, ...]]:
|
921 |
+
if cross_attention_kwargs is not None:
|
922 |
+
if cross_attention_kwargs.get("scale", None) is not None:
|
923 |
+
logger.warning("Passing `scale` to `cross_attention_kwargs` is deprecated. `scale` will be ignored.")
|
924 |
+
|
925 |
+
|
926 |
+
|
927 |
+
output_states = ()
|
928 |
+
|
929 |
+
blocks = list(zip(self.resnets, self.attentions))
|
930 |
+
|
931 |
+
for i, (resnet, attn) in enumerate(blocks):
|
932 |
+
if self.training and self.gradient_checkpointing:
|
933 |
+
|
934 |
+
def create_custom_forward(module, return_dict=None):
|
935 |
+
def custom_forward(*inputs):
|
936 |
+
if return_dict is not None:
|
937 |
+
return module(*inputs, return_dict=return_dict)
|
938 |
+
else:
|
939 |
+
return module(*inputs)
|
940 |
+
|
941 |
+
return custom_forward
|
942 |
+
|
943 |
+
ckpt_kwargs: Dict[str, Any] = {"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
|
944 |
+
hidden_states = torch.utils.checkpoint.checkpoint(
|
945 |
+
create_custom_forward(resnet),
|
946 |
+
hidden_states,
|
947 |
+
temb,
|
948 |
+
**ckpt_kwargs,
|
949 |
+
)
|
950 |
+
hidden_states = attn(
|
951 |
+
hidden_states,
|
952 |
+
encoder_hidden_states=encoder_hidden_states,
|
953 |
+
cross_attention_kwargs=cross_attention_kwargs,
|
954 |
+
attention_mask=attention_mask,
|
955 |
+
encoder_attention_mask=encoder_attention_mask,
|
956 |
+
return_dict=False,
|
957 |
+
)[0]
|
958 |
+
else:
|
959 |
+
hidden_states = resnet(hidden_states, temb)
|
960 |
+
hidden_states = attn(
|
961 |
+
hidden_states,
|
962 |
+
encoder_hidden_states=encoder_hidden_states,
|
963 |
+
cross_attention_kwargs=cross_attention_kwargs,
|
964 |
+
attention_mask=attention_mask,
|
965 |
+
encoder_attention_mask=encoder_attention_mask,
|
966 |
+
return_dict=False,
|
967 |
+
)[0]
|
968 |
+
|
969 |
+
hidden_states += additional_residuals.pop(0)
|
970 |
+
|
971 |
+
output_states = output_states + (hidden_states,)
|
972 |
+
|
973 |
+
if self.downsamplers is not None:
|
974 |
+
for downsampler in self.downsamplers:
|
975 |
+
hidden_states = downsampler(hidden_states)
|
976 |
+
hidden_states += additional_residuals.pop(0)
|
977 |
+
|
978 |
+
output_states = output_states + (hidden_states,)
|
979 |
+
|
980 |
+
return hidden_states, output_states
|
981 |
+
|
982 |
+
|
983 |
+
down_intrablock_additional_residuals = dino_down_block_additional_residuals
|
984 |
+
|
985 |
+
sample += down_intrablock_additional_residuals.pop(0)
|
986 |
+
down_block_res_samples = (sample,)
|
987 |
+
|
988 |
+
for downsample_block in self.down_blocks:
|
989 |
+
if hasattr(downsample_block, "has_cross_attention") and downsample_block.has_cross_attention:
|
990 |
+
|
991 |
+
sample, res_samples = residual_blockforward(
|
992 |
+
downsample_block,
|
993 |
+
hidden_states=sample,
|
994 |
+
temb=emb,
|
995 |
+
encoder_hidden_states=encoder_hidden_states,
|
996 |
+
attention_mask=attention_mask,
|
997 |
+
cross_attention_kwargs=cross_attention_kwargs,
|
998 |
+
encoder_attention_mask=encoder_attention_mask,
|
999 |
+
additional_residuals = down_intrablock_additional_residuals,
|
1000 |
+
)
|
1001 |
+
|
1002 |
+
else:
|
1003 |
+
sample, res_samples = residual_downforward(
|
1004 |
+
downsample_block,
|
1005 |
+
hidden_states=sample,
|
1006 |
+
temb=emb,
|
1007 |
+
additional_residuals = down_intrablock_additional_residuals,
|
1008 |
+
)
|
1009 |
+
|
1010 |
+
|
1011 |
+
down_block_res_samples += res_samples
|
1012 |
+
|
1013 |
+
|
1014 |
+
if is_controlnet:
|
1015 |
+
new_down_block_res_samples = ()
|
1016 |
+
|
1017 |
+
for down_block_res_sample, down_block_additional_residual in zip(
|
1018 |
+
down_block_res_samples, down_block_additional_residuals
|
1019 |
+
):
|
1020 |
+
down_block_res_sample = down_block_res_sample + down_block_additional_residual
|
1021 |
+
new_down_block_res_samples = new_down_block_res_samples + (down_block_res_sample,)
|
1022 |
+
|
1023 |
+
down_block_res_samples = new_down_block_res_samples
|
1024 |
+
|
1025 |
+
# 4. mid
|
1026 |
+
if self.mid_block is not None:
|
1027 |
+
if hasattr(self.mid_block, "has_cross_attention") and self.mid_block.has_cross_attention:
|
1028 |
+
sample = self.mid_block(
|
1029 |
+
sample,
|
1030 |
+
emb,
|
1031 |
+
encoder_hidden_states=encoder_hidden_states,
|
1032 |
+
attention_mask=attention_mask,
|
1033 |
+
cross_attention_kwargs=cross_attention_kwargs,
|
1034 |
+
encoder_attention_mask=encoder_attention_mask,
|
1035 |
+
)
|
1036 |
+
else:
|
1037 |
+
sample = self.mid_block(sample, emb)
|
1038 |
+
|
1039 |
+
# To support T2I-Adapter-XL
|
1040 |
+
if (
|
1041 |
+
is_adapter
|
1042 |
+
and len(down_intrablock_additional_residuals) > 0
|
1043 |
+
and sample.shape == down_intrablock_additional_residuals[0].shape
|
1044 |
+
):
|
1045 |
+
sample += down_intrablock_additional_residuals.pop(0)
|
1046 |
+
|
1047 |
+
if is_controlnet:
|
1048 |
+
sample = sample + mid_block_additional_residual
|
1049 |
+
|
1050 |
+
# 5. up
|
1051 |
+
for i, upsample_block in enumerate(self.up_blocks):
|
1052 |
+
is_final_block = i == len(self.up_blocks) - 1
|
1053 |
+
|
1054 |
+
res_samples = down_block_res_samples[-len(upsample_block.resnets) :]
|
1055 |
+
down_block_res_samples = down_block_res_samples[: -len(upsample_block.resnets)]
|
1056 |
+
|
1057 |
+
# if we have not reached the final block and need to forward the
|
1058 |
+
# upsample size, we do it here
|
1059 |
+
if not is_final_block and forward_upsample_size:
|
1060 |
+
upsample_size = down_block_res_samples[-1].shape[2:]
|
1061 |
+
|
1062 |
+
if hasattr(upsample_block, "has_cross_attention") and upsample_block.has_cross_attention:
|
1063 |
+
sample = upsample_block(
|
1064 |
+
hidden_states=sample,
|
1065 |
+
temb=emb,
|
1066 |
+
res_hidden_states_tuple=res_samples,
|
1067 |
+
encoder_hidden_states=encoder_hidden_states,
|
1068 |
+
cross_attention_kwargs=cross_attention_kwargs,
|
1069 |
+
upsample_size=upsample_size,
|
1070 |
+
attention_mask=attention_mask,
|
1071 |
+
encoder_attention_mask=encoder_attention_mask,
|
1072 |
+
)
|
1073 |
+
else:
|
1074 |
+
sample = upsample_block(
|
1075 |
+
hidden_states=sample,
|
1076 |
+
temb=emb,
|
1077 |
+
res_hidden_states_tuple=res_samples,
|
1078 |
+
upsample_size=upsample_size,
|
1079 |
+
)
|
1080 |
+
|
1081 |
+
# 6. post-process
|
1082 |
+
if self.conv_norm_out:
|
1083 |
+
sample = self.conv_norm_out(sample)
|
1084 |
+
sample = self.conv_act(sample)
|
1085 |
+
sample = self.conv_out(sample)
|
1086 |
+
|
1087 |
+
if USE_PEFT_BACKEND:
|
1088 |
+
# remove `lora_scale` from each PEFT layer
|
1089 |
+
unscale_lora_layers(self, lora_scale)
|
1090 |
+
|
1091 |
+
if not return_dict:
|
1092 |
+
return (sample,)
|
1093 |
+
|
1094 |
+
return UNet2DConditionOutput(sample=sample)
|
1095 |
+
|
1096 |
+
|
1097 |
+
|
1098 |
+
@staticmethod
|
1099 |
+
def ensemble_normals(
|
1100 |
+
normals: torch.Tensor, output_uncertainty: bool, reduction: str = "closest"
|
1101 |
+
) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
|
1102 |
+
"""
|
1103 |
+
Ensembles the normals maps represented by the `normals` tensor with expected shape `(B, 3, H, W)`, where B is
|
1104 |
+
the number of ensemble members for a given prediction of size `(H x W)`.
|
1105 |
+
|
1106 |
+
Args:
|
1107 |
+
normals (`torch.Tensor`):
|
1108 |
+
Input ensemble normals maps.
|
1109 |
+
output_uncertainty (`bool`, *optional*, defaults to `False`):
|
1110 |
+
Whether to output uncertainty map.
|
1111 |
+
reduction (`str`, *optional*, defaults to `"closest"`):
|
1112 |
+
Reduction method used to ensemble aligned predictions. The accepted values are: `"closest"` and
|
1113 |
+
`"mean"`.
|
1114 |
+
|
1115 |
+
Returns:
|
1116 |
+
A tensor of aligned and ensembled normals maps with shape `(1, 3, H, W)` and optionally a tensor of
|
1117 |
+
uncertainties of shape `(1, 1, H, W)`.
|
1118 |
+
"""
|
1119 |
+
if normals.dim() != 4 or normals.shape[1] != 3:
|
1120 |
+
raise ValueError(f"Expecting 4D tensor of shape [B,3,H,W]; got {normals.shape}.")
|
1121 |
+
if reduction not in ("closest", "mean"):
|
1122 |
+
raise ValueError(f"Unrecognized reduction method: {reduction}.")
|
1123 |
+
|
1124 |
+
mean_normals = normals.mean(dim=0, keepdim=True) # [1,3,H,W]
|
1125 |
+
mean_normals = MarigoldNormalsPipeline.normalize_normals(mean_normals) # [1,3,H,W]
|
1126 |
+
|
1127 |
+
sim_cos = (mean_normals * normals).sum(dim=1, keepdim=True) # [E,1,H,W]
|
1128 |
+
sim_cos = sim_cos.clamp(-1, 1) # required to avoid NaN in uncertainty with fp16
|
1129 |
+
|
1130 |
+
uncertainty = None
|
1131 |
+
if output_uncertainty:
|
1132 |
+
uncertainty = sim_cos.arccos() # [E,1,H,W]
|
1133 |
+
uncertainty = uncertainty.mean(dim=0, keepdim=True) / np.pi # [1,1,H,W]
|
1134 |
+
|
1135 |
+
if reduction == "mean":
|
1136 |
+
return mean_normals, uncertainty # [1,3,H,W], [1,1,H,W]
|
1137 |
+
|
1138 |
+
closest_indices = sim_cos.argmax(dim=0, keepdim=True) # [1,1,H,W]
|
1139 |
+
closest_indices = closest_indices.repeat(1, 3, 1, 1) # [1,3,H,W]
|
1140 |
+
closest_normals = torch.gather(normals, 0, closest_indices) # [1,3,H,W]
|
1141 |
+
|
1142 |
+
return closest_normals, uncertainty # [1,3,H,W], [1,1,H,W]
|
1143 |
+
|
1144 |
+
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
|
1145 |
+
def retrieve_timesteps(
|
1146 |
+
scheduler,
|
1147 |
+
num_inference_steps: Optional[int] = None,
|
1148 |
+
device: Optional[Union[str, torch.device]] = None,
|
1149 |
+
timesteps: Optional[List[int]] = None,
|
1150 |
+
sigmas: Optional[List[float]] = None,
|
1151 |
+
**kwargs,
|
1152 |
+
):
|
1153 |
+
"""
|
1154 |
+
Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
|
1155 |
+
custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
|
1156 |
+
|
1157 |
+
Args:
|
1158 |
+
scheduler (`SchedulerMixin`):
|
1159 |
+
The scheduler to get timesteps from.
|
1160 |
+
num_inference_steps (`int`):
|
1161 |
+
The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
|
1162 |
+
must be `None`.
|
1163 |
+
device (`str` or `torch.device`, *optional*):
|
1164 |
+
The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
|
1165 |
+
timesteps (`List[int]`, *optional*):
|
1166 |
+
Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
|
1167 |
+
`num_inference_steps` and `sigmas` must be `None`.
|
1168 |
+
sigmas (`List[float]`, *optional*):
|
1169 |
+
Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
|
1170 |
+
`num_inference_steps` and `timesteps` must be `None`.
|
1171 |
+
|
1172 |
+
Returns:
|
1173 |
+
`Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
|
1174 |
+
second element is the number of inference steps.
|
1175 |
+
"""
|
1176 |
+
if timesteps is not None and sigmas is not None:
|
1177 |
+
raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
|
1178 |
+
if timesteps is not None:
|
1179 |
+
accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
|
1180 |
+
if not accepts_timesteps:
|
1181 |
+
raise ValueError(
|
1182 |
+
f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
|
1183 |
+
f" timestep schedules. Please check whether you are using the correct scheduler."
|
1184 |
+
)
|
1185 |
+
scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
|
1186 |
+
timesteps = scheduler.timesteps
|
1187 |
+
num_inference_steps = len(timesteps)
|
1188 |
+
elif sigmas is not None:
|
1189 |
+
accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
|
1190 |
+
if not accept_sigmas:
|
1191 |
+
raise ValueError(
|
1192 |
+
f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
|
1193 |
+
f" sigmas schedules. Please check whether you are using the correct scheduler."
|
1194 |
+
)
|
1195 |
+
scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
|
1196 |
+
timesteps = scheduler.timesteps
|
1197 |
+
num_inference_steps = len(timesteps)
|
1198 |
+
else:
|
1199 |
+
scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
|
1200 |
+
timesteps = scheduler.timesteps
|
1201 |
+
return timesteps, num_inference_steps
|
stablenormal/pipeline_yoso_normal.py
ADDED
@@ -0,0 +1,723 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Copyright 2024 Marigold authors, PRS ETH Zurich. All rights reserved.
|
2 |
+
# Copyright 2024 The HuggingFace Team. All rights reserved.
|
3 |
+
#
|
4 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
5 |
+
# you may not use this file except in compliance with the License.
|
6 |
+
# You may obtain a copy of the License at
|
7 |
+
#
|
8 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
9 |
+
#
|
10 |
+
# Unless required by applicable law or agreed to in writing, software
|
11 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
12 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
13 |
+
# See the License for the specific language governing permissions and
|
14 |
+
# limitations under the License.
|
15 |
+
# --------------------------------------------------------------------------
|
16 |
+
# More information and citation instructions are available on the
|
17 |
+
# --------------------------------------------------------------------------
|
18 |
+
from dataclasses import dataclass
|
19 |
+
from typing import Any, Dict, List, Optional, Tuple, Union
|
20 |
+
|
21 |
+
import numpy as np
|
22 |
+
import torch
|
23 |
+
from PIL import Image
|
24 |
+
from tqdm.auto import tqdm
|
25 |
+
from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
|
26 |
+
|
27 |
+
|
28 |
+
from diffusers.image_processor import PipelineImageInput
|
29 |
+
from diffusers.models import (
|
30 |
+
AutoencoderKL,
|
31 |
+
UNet2DConditionModel,
|
32 |
+
ControlNetModel,
|
33 |
+
)
|
34 |
+
from diffusers.schedulers import (
|
35 |
+
DDIMScheduler
|
36 |
+
)
|
37 |
+
|
38 |
+
from diffusers.utils import (
|
39 |
+
BaseOutput,
|
40 |
+
logging,
|
41 |
+
replace_example_docstring,
|
42 |
+
)
|
43 |
+
|
44 |
+
|
45 |
+
from diffusers.utils.torch_utils import randn_tensor
|
46 |
+
from diffusers.pipelines.controlnet import StableDiffusionControlNetPipeline
|
47 |
+
from diffusers.pipelines.pipeline_utils import DiffusionPipeline
|
48 |
+
from diffusers.pipelines.marigold.marigold_image_processing import MarigoldImageProcessor
|
49 |
+
from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
|
50 |
+
|
51 |
+
import pdb
|
52 |
+
|
53 |
+
|
54 |
+
|
55 |
+
logger = logging.get_logger(__name__) # pylint: disable=invalid-name
|
56 |
+
|
57 |
+
|
58 |
+
EXAMPLE_DOC_STRING = """
|
59 |
+
Examples:
|
60 |
+
```py
|
61 |
+
>>> import diffusers
|
62 |
+
>>> import torch
|
63 |
+
|
64 |
+
>>> pipe = diffusers.MarigoldNormalsPipeline.from_pretrained(
|
65 |
+
... "prs-eth/marigold-normals-lcm-v0-1", variant="fp16", torch_dtype=torch.float16
|
66 |
+
... ).to("cuda")
|
67 |
+
|
68 |
+
>>> image = diffusers.utils.load_image("https://marigoldmonodepth.github.io/images/einstein.jpg")
|
69 |
+
>>> normals = pipe(image)
|
70 |
+
|
71 |
+
>>> vis = pipe.image_processor.visualize_normals(normals.prediction)
|
72 |
+
>>> vis[0].save("einstein_normals.png")
|
73 |
+
```
|
74 |
+
"""
|
75 |
+
|
76 |
+
|
77 |
+
@dataclass
|
78 |
+
class YosoNormalsOutput(BaseOutput):
|
79 |
+
"""
|
80 |
+
Output class for Marigold monocular normals prediction pipeline.
|
81 |
+
|
82 |
+
Args:
|
83 |
+
prediction (`np.ndarray`, `torch.Tensor`):
|
84 |
+
Predicted normals with values in the range [-1, 1]. The shape is always $numimages \times 3 \times height
|
85 |
+
\times width$, regardless of whether the images were passed as a 4D array or a list.
|
86 |
+
uncertainty (`None`, `np.ndarray`, `torch.Tensor`):
|
87 |
+
Uncertainty maps computed from the ensemble, with values in the range [0, 1]. The shape is $numimages
|
88 |
+
\times 1 \times height \times width$.
|
89 |
+
latent (`None`, `torch.Tensor`):
|
90 |
+
Latent features corresponding to the predictions, compatible with the `latents` argument of the pipeline.
|
91 |
+
The shape is $numimages * numensemble \times 4 \times latentheight \times latentwidth$.
|
92 |
+
"""
|
93 |
+
|
94 |
+
prediction: Union[np.ndarray, torch.Tensor]
|
95 |
+
latent: Union[None, torch.Tensor]
|
96 |
+
gauss_latent: Union[None, torch.Tensor]
|
97 |
+
|
98 |
+
|
99 |
+
class YOSONormalsPipeline(StableDiffusionControlNetPipeline):
|
100 |
+
""" Pipeline for monocular normals estimation using the Marigold method: https://marigoldmonodepth.github.io.
|
101 |
+
Pipeline for text-to-image generation using Stable Diffusion with ControlNet guidance.
|
102 |
+
|
103 |
+
This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
|
104 |
+
implemented for all pipelines (downloading, saving, running on a particular device, etc.).
|
105 |
+
|
106 |
+
The pipeline also inherits the following loading methods:
|
107 |
+
- [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
|
108 |
+
- [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
|
109 |
+
- [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
|
110 |
+
- [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
|
111 |
+
- [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
|
112 |
+
|
113 |
+
Args:
|
114 |
+
vae ([`AutoencoderKL`]):
|
115 |
+
Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations.
|
116 |
+
text_encoder ([`~transformers.CLIPTextModel`]):
|
117 |
+
Frozen text-encoder ([clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14)).
|
118 |
+
tokenizer ([`~transformers.CLIPTokenizer`]):
|
119 |
+
A `CLIPTokenizer` to tokenize text.
|
120 |
+
unet ([`UNet2DConditionModel`]):
|
121 |
+
A `UNet2DConditionModel` to denoise the encoded image latents.
|
122 |
+
controlnet ([`ControlNetModel`] or `List[ControlNetModel]`):
|
123 |
+
Provides additional conditioning to the `unet` during the denoising process. If you set multiple
|
124 |
+
ControlNets as a list, the outputs from each ControlNet are added together to create one combined
|
125 |
+
additional conditioning.
|
126 |
+
scheduler ([`SchedulerMixin`]):
|
127 |
+
A scheduler to be used in combination with `unet` to denoise the encoded image latents. Can be one of
|
128 |
+
[`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
|
129 |
+
safety_checker ([`StableDiffusionSafetyChecker`]):
|
130 |
+
Classification module that estimates whether generated images could be considered offensive or harmful.
|
131 |
+
Please refer to the [model card](https://huggingface.co/runwayml/stable-diffusion-v1-5) for more details
|
132 |
+
about a model's potential harms.
|
133 |
+
feature_extractor ([`~transformers.CLIPImageProcessor`]):
|
134 |
+
A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
|
135 |
+
"""
|
136 |
+
|
137 |
+
model_cpu_offload_seq = "text_encoder->image_encoder->unet->vae"
|
138 |
+
_optional_components = ["safety_checker", "feature_extractor", "image_encoder"]
|
139 |
+
_exclude_from_cpu_offload = ["safety_checker"]
|
140 |
+
_callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"]
|
141 |
+
|
142 |
+
|
143 |
+
|
144 |
+
def __init__(
|
145 |
+
self,
|
146 |
+
vae: AutoencoderKL,
|
147 |
+
text_encoder: CLIPTextModel,
|
148 |
+
tokenizer: CLIPTokenizer,
|
149 |
+
unet: UNet2DConditionModel,
|
150 |
+
controlnet: Union[ControlNetModel, List[ControlNetModel], Tuple[ControlNetModel]],
|
151 |
+
scheduler: Union[DDIMScheduler],
|
152 |
+
safety_checker: StableDiffusionSafetyChecker,
|
153 |
+
feature_extractor: CLIPImageProcessor,
|
154 |
+
image_encoder: CLIPVisionModelWithProjection = None,
|
155 |
+
requires_safety_checker: bool = True,
|
156 |
+
default_denoising_steps: Optional[int] = 1,
|
157 |
+
default_processing_resolution: Optional[int] = 768,
|
158 |
+
prompt="",
|
159 |
+
empty_text_embedding=None,
|
160 |
+
t_start: Optional[int] = 401,
|
161 |
+
):
|
162 |
+
super().__init__(
|
163 |
+
vae,
|
164 |
+
text_encoder,
|
165 |
+
tokenizer,
|
166 |
+
unet,
|
167 |
+
controlnet,
|
168 |
+
scheduler,
|
169 |
+
safety_checker,
|
170 |
+
feature_extractor,
|
171 |
+
image_encoder,
|
172 |
+
requires_safety_checker,
|
173 |
+
)
|
174 |
+
|
175 |
+
# TODO yoso ImageProcessor
|
176 |
+
self.image_processor = MarigoldImageProcessor(vae_scale_factor=self.vae_scale_factor)
|
177 |
+
self.control_image_processor = MarigoldImageProcessor(vae_scale_factor=self.vae_scale_factor)
|
178 |
+
self.default_denoising_steps = default_denoising_steps
|
179 |
+
self.default_processing_resolution = default_processing_resolution
|
180 |
+
self.prompt = prompt
|
181 |
+
self.prompt_embeds = None
|
182 |
+
self.empty_text_embedding = empty_text_embedding
|
183 |
+
self.t_start= t_start # target_out latents
|
184 |
+
self.gauss_latent = None
|
185 |
+
|
186 |
+
def check_inputs(
|
187 |
+
self,
|
188 |
+
image: PipelineImageInput,
|
189 |
+
num_inference_steps: int,
|
190 |
+
ensemble_size: int,
|
191 |
+
processing_resolution: int,
|
192 |
+
resample_method_input: str,
|
193 |
+
resample_method_output: str,
|
194 |
+
batch_size: int,
|
195 |
+
ensembling_kwargs: Optional[Dict[str, Any]],
|
196 |
+
latents: Optional[torch.Tensor],
|
197 |
+
generator: Optional[Union[torch.Generator, List[torch.Generator]]],
|
198 |
+
output_type: str,
|
199 |
+
output_uncertainty: bool,
|
200 |
+
) -> int:
|
201 |
+
if num_inference_steps is None:
|
202 |
+
raise ValueError("`num_inference_steps` is not specified and could not be resolved from the model config.")
|
203 |
+
if num_inference_steps < 1:
|
204 |
+
raise ValueError("`num_inference_steps` must be positive.")
|
205 |
+
if ensemble_size < 1:
|
206 |
+
raise ValueError("`ensemble_size` must be positive.")
|
207 |
+
if ensemble_size == 2:
|
208 |
+
logger.warning(
|
209 |
+
"`ensemble_size` == 2 results are similar to no ensembling (1); "
|
210 |
+
"consider increasing the value to at least 3."
|
211 |
+
)
|
212 |
+
if ensemble_size == 1 and output_uncertainty:
|
213 |
+
raise ValueError(
|
214 |
+
"Computing uncertainty by setting `output_uncertainty=True` also requires setting `ensemble_size` "
|
215 |
+
"greater than 1."
|
216 |
+
)
|
217 |
+
if processing_resolution is None:
|
218 |
+
raise ValueError(
|
219 |
+
"`processing_resolution` is not specified and could not be resolved from the model config."
|
220 |
+
)
|
221 |
+
if processing_resolution < 0:
|
222 |
+
raise ValueError(
|
223 |
+
"`processing_resolution` must be non-negative: 0 for native resolution, or any positive value for "
|
224 |
+
"downsampled processing."
|
225 |
+
)
|
226 |
+
if processing_resolution % self.vae_scale_factor != 0:
|
227 |
+
raise ValueError(f"`processing_resolution` must be a multiple of {self.vae_scale_factor}.")
|
228 |
+
if resample_method_input not in ("nearest", "nearest-exact", "bilinear", "bicubic", "area"):
|
229 |
+
raise ValueError(
|
230 |
+
"`resample_method_input` takes string values compatible with PIL library: "
|
231 |
+
"nearest, nearest-exact, bilinear, bicubic, area."
|
232 |
+
)
|
233 |
+
if resample_method_output not in ("nearest", "nearest-exact", "bilinear", "bicubic", "area"):
|
234 |
+
raise ValueError(
|
235 |
+
"`resample_method_output` takes string values compatible with PIL library: "
|
236 |
+
"nearest, nearest-exact, bilinear, bicubic, area."
|
237 |
+
)
|
238 |
+
if batch_size < 1:
|
239 |
+
raise ValueError("`batch_size` must be positive.")
|
240 |
+
if output_type not in ["pt", "np"]:
|
241 |
+
raise ValueError("`output_type` must be one of `pt` or `np`.")
|
242 |
+
if latents is not None and generator is not None:
|
243 |
+
raise ValueError("`latents` and `generator` cannot be used together.")
|
244 |
+
if ensembling_kwargs is not None:
|
245 |
+
if not isinstance(ensembling_kwargs, dict):
|
246 |
+
raise ValueError("`ensembling_kwargs` must be a dictionary.")
|
247 |
+
if "reduction" in ensembling_kwargs and ensembling_kwargs["reduction"] not in ("closest", "mean"):
|
248 |
+
raise ValueError("`ensembling_kwargs['reduction']` can be either `'closest'` or `'mean'`.")
|
249 |
+
|
250 |
+
# image checks
|
251 |
+
num_images = 0
|
252 |
+
W, H = None, None
|
253 |
+
if not isinstance(image, list):
|
254 |
+
image = [image]
|
255 |
+
for i, img in enumerate(image):
|
256 |
+
if isinstance(img, np.ndarray) or torch.is_tensor(img):
|
257 |
+
if img.ndim not in (2, 3, 4):
|
258 |
+
raise ValueError(f"`image[{i}]` has unsupported dimensions or shape: {img.shape}.")
|
259 |
+
H_i, W_i = img.shape[-2:]
|
260 |
+
N_i = 1
|
261 |
+
if img.ndim == 4:
|
262 |
+
N_i = img.shape[0]
|
263 |
+
elif isinstance(img, Image.Image):
|
264 |
+
W_i, H_i = img.size
|
265 |
+
N_i = 1
|
266 |
+
else:
|
267 |
+
raise ValueError(f"Unsupported `image[{i}]` type: {type(img)}.")
|
268 |
+
if W is None:
|
269 |
+
W, H = W_i, H_i
|
270 |
+
elif (W, H) != (W_i, H_i):
|
271 |
+
raise ValueError(
|
272 |
+
f"Input `image[{i}]` has incompatible dimensions {(W_i, H_i)} with the previous images {(W, H)}"
|
273 |
+
)
|
274 |
+
num_images += N_i
|
275 |
+
|
276 |
+
# latents checks
|
277 |
+
if latents is not None:
|
278 |
+
if not torch.is_tensor(latents):
|
279 |
+
raise ValueError("`latents` must be a torch.Tensor.")
|
280 |
+
if latents.dim() != 4:
|
281 |
+
raise ValueError(f"`latents` has unsupported dimensions or shape: {latents.shape}.")
|
282 |
+
|
283 |
+
if processing_resolution > 0:
|
284 |
+
max_orig = max(H, W)
|
285 |
+
new_H = H * processing_resolution // max_orig
|
286 |
+
new_W = W * processing_resolution // max_orig
|
287 |
+
if new_H == 0 or new_W == 0:
|
288 |
+
raise ValueError(f"Extreme aspect ratio of the input image: [{W} x {H}]")
|
289 |
+
W, H = new_W, new_H
|
290 |
+
w = (W + self.vae_scale_factor - 1) // self.vae_scale_factor
|
291 |
+
h = (H + self.vae_scale_factor - 1) // self.vae_scale_factor
|
292 |
+
shape_expected = (num_images * ensemble_size, self.vae.config.latent_channels, h, w)
|
293 |
+
|
294 |
+
if latents.shape != shape_expected:
|
295 |
+
raise ValueError(f"`latents` has unexpected shape={latents.shape} expected={shape_expected}.")
|
296 |
+
|
297 |
+
# generator checks
|
298 |
+
if generator is not None:
|
299 |
+
if isinstance(generator, list):
|
300 |
+
if len(generator) != num_images * ensemble_size:
|
301 |
+
raise ValueError(
|
302 |
+
"The number of generators must match the total number of ensemble members for all input images."
|
303 |
+
)
|
304 |
+
if not all(g.device.type == generator[0].device.type for g in generator):
|
305 |
+
raise ValueError("`generator` device placement is not consistent in the list.")
|
306 |
+
elif not isinstance(generator, torch.Generator):
|
307 |
+
raise ValueError(f"Unsupported generator type: {type(generator)}.")
|
308 |
+
|
309 |
+
return num_images
|
310 |
+
|
311 |
+
def progress_bar(self, iterable=None, total=None, desc=None, leave=True):
|
312 |
+
if not hasattr(self, "_progress_bar_config"):
|
313 |
+
self._progress_bar_config = {}
|
314 |
+
elif not isinstance(self._progress_bar_config, dict):
|
315 |
+
raise ValueError(
|
316 |
+
f"`self._progress_bar_config` should be of type `dict`, but is {type(self._progress_bar_config)}."
|
317 |
+
)
|
318 |
+
|
319 |
+
progress_bar_config = dict(**self._progress_bar_config)
|
320 |
+
progress_bar_config["desc"] = progress_bar_config.get("desc", desc)
|
321 |
+
progress_bar_config["leave"] = progress_bar_config.get("leave", leave)
|
322 |
+
if iterable is not None:
|
323 |
+
return tqdm(iterable, **progress_bar_config)
|
324 |
+
elif total is not None:
|
325 |
+
return tqdm(total=total, **progress_bar_config)
|
326 |
+
else:
|
327 |
+
raise ValueError("Either `total` or `iterable` has to be defined.")
|
328 |
+
|
329 |
+
@torch.no_grad()
|
330 |
+
@replace_example_docstring(EXAMPLE_DOC_STRING)
|
331 |
+
def __call__(
|
332 |
+
self,
|
333 |
+
image: PipelineImageInput,
|
334 |
+
prompt: Union[str, List[str]] = None,
|
335 |
+
negative_prompt: Optional[Union[str, List[str]]] = None,
|
336 |
+
num_inference_steps: Optional[int] = None,
|
337 |
+
ensemble_size: int = 1,
|
338 |
+
processing_resolution: Optional[int] = None,
|
339 |
+
match_input_resolution: bool = True,
|
340 |
+
resample_method_input: str = "bilinear",
|
341 |
+
resample_method_output: str = "bilinear",
|
342 |
+
batch_size: int = 1,
|
343 |
+
ensembling_kwargs: Optional[Dict[str, Any]] = None,
|
344 |
+
latents: Optional[Union[torch.Tensor, List[torch.Tensor]]] = None,
|
345 |
+
prompt_embeds: Optional[torch.Tensor] = None,
|
346 |
+
negative_prompt_embeds: Optional[torch.Tensor] = None,
|
347 |
+
num_images_per_prompt: Optional[int] = 1,
|
348 |
+
generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
|
349 |
+
controlnet_conditioning_scale: Union[float, List[float]] = 1.0,
|
350 |
+
output_type: str = "np",
|
351 |
+
output_uncertainty: bool = False,
|
352 |
+
output_latent: bool = False,
|
353 |
+
skip_preprocess: bool = False,
|
354 |
+
return_dict: bool = True,
|
355 |
+
**kwargs,
|
356 |
+
):
|
357 |
+
"""
|
358 |
+
Function invoked when calling the pipeline.
|
359 |
+
|
360 |
+
Args:
|
361 |
+
image (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`),
|
362 |
+
`List[torch.Tensor]`: An input image or images used as an input for the normals estimation task. For
|
363 |
+
arrays and tensors, the expected value range is between `[0, 1]`. Passing a batch of images is possible
|
364 |
+
by providing a four-dimensional array or a tensor. Additionally, a list of images of two- or
|
365 |
+
three-dimensional arrays or tensors can be passed. In the latter case, all list elements must have the
|
366 |
+
same width and height.
|
367 |
+
num_inference_steps (`int`, *optional*, defaults to `None`):
|
368 |
+
Number of denoising diffusion steps during inference. The default value `None` results in automatic
|
369 |
+
selection. The number of steps should be at least 10 with the full Marigold models, and between 1 and 4
|
370 |
+
for Marigold-LCM models.
|
371 |
+
ensemble_size (`int`, defaults to `1`):
|
372 |
+
Number of ensemble predictions. Recommended values are 5 and higher for better precision, or 1 for
|
373 |
+
faster inference.
|
374 |
+
processing_resolution (`int`, *optional*, defaults to `None`):
|
375 |
+
Effective processing resolution. When set to `0`, matches the larger input image dimension. This
|
376 |
+
produces crisper predictions, but may also lead to the overall loss of global context. The default
|
377 |
+
value `None` resolves to the optimal value from the model config.
|
378 |
+
match_input_resolution (`bool`, *optional*, defaults to `True`):
|
379 |
+
When enabled, the output prediction is resized to match the input dimensions. When disabled, the longer
|
380 |
+
side of the output will equal to `processing_resolution`.
|
381 |
+
resample_method_input (`str`, *optional*, defaults to `"bilinear"`):
|
382 |
+
Resampling method used to resize input images to `processing_resolution`. The accepted values are:
|
383 |
+
`"nearest"`, `"nearest-exact"`, `"bilinear"`, `"bicubic"`, or `"area"`.
|
384 |
+
resample_method_output (`str`, *optional*, defaults to `"bilinear"`):
|
385 |
+
Resampling method used to resize output predictions to match the input resolution. The accepted values
|
386 |
+
are `"nearest"`, `"nearest-exact"`, `"bilinear"`, `"bicubic"`, or `"area"`.
|
387 |
+
batch_size (`int`, *optional*, defaults to `1`):
|
388 |
+
Batch size; only matters when setting `ensemble_size` or passing a tensor of images.
|
389 |
+
ensembling_kwargs (`dict`, *optional*, defaults to `None`)
|
390 |
+
Extra dictionary with arguments for precise ensembling control. The following options are available:
|
391 |
+
- reduction (`str`, *optional*, defaults to `"closest"`): Defines the ensembling function applied in
|
392 |
+
every pixel location, can be either `"closest"` or `"mean"`.
|
393 |
+
latents (`torch.Tensor`, *optional*, defaults to `None`):
|
394 |
+
Latent noise tensors to replace the random initialization. These can be taken from the previous
|
395 |
+
function call's output.
|
396 |
+
generator (`torch.Generator`, or `List[torch.Generator]`, *optional*, defaults to `None`):
|
397 |
+
Random number generator object to ensure reproducibility.
|
398 |
+
output_type (`str`, *optional*, defaults to `"np"`):
|
399 |
+
Preferred format of the output's `prediction` and the optional `uncertainty` fields. The accepted
|
400 |
+
values are: `"np"` (numpy array) or `"pt"` (torch tensor).
|
401 |
+
output_uncertainty (`bool`, *optional*, defaults to `False`):
|
402 |
+
When enabled, the output's `uncertainty` field contains the predictive uncertainty map, provided that
|
403 |
+
the `ensemble_size` argument is set to a value above 2.
|
404 |
+
output_latent (`bool`, *optional*, defaults to `False`):
|
405 |
+
When enabled, the output's `latent` field contains the latent codes corresponding to the predictions
|
406 |
+
within the ensemble. These codes can be saved, modified, and used for subsequent calls with the
|
407 |
+
`latents` argument.
|
408 |
+
return_dict (`bool`, *optional*, defaults to `True`):
|
409 |
+
Whether or not to return a [`~pipelines.marigold.MarigoldDepthOutput`] instead of a plain tuple.
|
410 |
+
|
411 |
+
Examples:
|
412 |
+
|
413 |
+
Returns:
|
414 |
+
[`~pipelines.marigold.MarigoldNormalsOutput`] or `tuple`:
|
415 |
+
If `return_dict` is `True`, [`~pipelines.marigold.MarigoldNormalsOutput`] is returned, otherwise a
|
416 |
+
`tuple` is returned where the first element is the prediction, the second element is the uncertainty
|
417 |
+
(or `None`), and the third is the latent (or `None`).
|
418 |
+
"""
|
419 |
+
|
420 |
+
# 0. Resolving variables.
|
421 |
+
device = self._execution_device
|
422 |
+
dtype = self.dtype
|
423 |
+
|
424 |
+
# Model-specific optimal default values leading to fast and reasonable results.
|
425 |
+
if num_inference_steps is None:
|
426 |
+
num_inference_steps = self.default_denoising_steps
|
427 |
+
if processing_resolution is None:
|
428 |
+
processing_resolution = self.default_processing_resolution
|
429 |
+
|
430 |
+
# 1. Check inputs.
|
431 |
+
num_images = self.check_inputs(
|
432 |
+
image,
|
433 |
+
num_inference_steps,
|
434 |
+
ensemble_size,
|
435 |
+
processing_resolution,
|
436 |
+
resample_method_input,
|
437 |
+
resample_method_output,
|
438 |
+
batch_size,
|
439 |
+
ensembling_kwargs,
|
440 |
+
latents,
|
441 |
+
generator,
|
442 |
+
output_type,
|
443 |
+
output_uncertainty,
|
444 |
+
)
|
445 |
+
|
446 |
+
|
447 |
+
# 2. Prepare empty text conditioning.
|
448 |
+
# Model invocation: self.tokenizer, self.text_encoder.
|
449 |
+
if self.empty_text_embedding is None:
|
450 |
+
prompt = ""
|
451 |
+
text_inputs = self.tokenizer(
|
452 |
+
prompt,
|
453 |
+
padding="do_not_pad",
|
454 |
+
max_length=self.tokenizer.model_max_length,
|
455 |
+
truncation=True,
|
456 |
+
return_tensors="pt",
|
457 |
+
)
|
458 |
+
text_input_ids = text_inputs.input_ids.to(device)
|
459 |
+
self.empty_text_embedding = self.text_encoder(text_input_ids)[0] # [1,2,1024]
|
460 |
+
|
461 |
+
|
462 |
+
|
463 |
+
# 3. prepare prompt
|
464 |
+
if self.prompt_embeds is None:
|
465 |
+
prompt_embeds, negative_prompt_embeds = self.encode_prompt(
|
466 |
+
self.prompt,
|
467 |
+
device,
|
468 |
+
num_images_per_prompt,
|
469 |
+
False,
|
470 |
+
negative_prompt,
|
471 |
+
prompt_embeds=prompt_embeds,
|
472 |
+
negative_prompt_embeds=None,
|
473 |
+
lora_scale=None,
|
474 |
+
clip_skip=None,
|
475 |
+
)
|
476 |
+
self.prompt_embeds = prompt_embeds
|
477 |
+
self.negative_prompt_embeds = negative_prompt_embeds
|
478 |
+
|
479 |
+
|
480 |
+
|
481 |
+
# 4. Preprocess input images. This function loads input image or images of compatible dimensions `(H, W)`,
|
482 |
+
# optionally downsamples them to the `processing_resolution` `(PH, PW)`, where
|
483 |
+
# `max(PH, PW) == processing_resolution`, and pads the dimensions to `(PPH, PPW)` such that these values are
|
484 |
+
# divisible by the latent space downscaling factor (typically 8 in Stable Diffusion). The default value `None`
|
485 |
+
# of `processing_resolution` resolves to the optimal value from the model config. It is a recommended mode of
|
486 |
+
# operation and leads to the most reasonable results. Using the native image resolution or any other processing
|
487 |
+
# resolution can lead to loss of either fine details or global context in the output predictions.
|
488 |
+
if not skip_preprocess:
|
489 |
+
image, padding, original_resolution = self.image_processor.preprocess(
|
490 |
+
image, processing_resolution, resample_method_input, device, dtype
|
491 |
+
) # [N,3,PPH,PPW]
|
492 |
+
else:
|
493 |
+
padding = (0, 0)
|
494 |
+
original_resolution = image.shape[2:]
|
495 |
+
# 5. Encode input image into latent space. At this step, each of the `N` input images is represented with `E`
|
496 |
+
# ensemble members. Each ensemble member is an independent diffused prediction, just initialized independently.
|
497 |
+
# Latents of each such predictions across all input images and all ensemble members are represented in the
|
498 |
+
# `pred_latent` variable. The variable `image_latent` is of the same shape: it contains each input image encoded
|
499 |
+
# into latent space and replicated `E` times. The latents can be either generated (see `generator` to ensure
|
500 |
+
# reproducibility), or passed explicitly via the `latents` argument. The latter can be set outside the pipeline
|
501 |
+
# code. For example, in the Marigold-LCM video processing demo, the latents initialization of a frame is taken
|
502 |
+
# as a convex combination of the latents output of the pipeline for the previous frame and a newly-sampled
|
503 |
+
# noise. This behavior can be achieved by setting the `output_latent` argument to `True`. The latent space
|
504 |
+
# dimensions are `(h, w)`. Encoding into latent space happens in batches of size `batch_size`.
|
505 |
+
# Model invocation: self.vae.encoder.
|
506 |
+
image_latent, gauss_latent = self.prepare_latents(
|
507 |
+
image, latents, generator, ensemble_size, batch_size
|
508 |
+
) # [N*E,4,h,w], [N*E,4,h,w]
|
509 |
+
|
510 |
+
del image
|
511 |
+
|
512 |
+
|
513 |
+
# 6. obtain control_output
|
514 |
+
|
515 |
+
cond_scale =controlnet_conditioning_scale
|
516 |
+
down_block_res_samples, mid_block_res_sample = self.controlnet(
|
517 |
+
image_latent.detach(),
|
518 |
+
self.t_start,
|
519 |
+
encoder_hidden_states=self.prompt_embeds,
|
520 |
+
conditioning_scale=cond_scale,
|
521 |
+
guess_mode=False,
|
522 |
+
return_dict=False,
|
523 |
+
)
|
524 |
+
|
525 |
+
|
526 |
+
# 7. YOSO sampling
|
527 |
+
latent_x_t = self.unet(
|
528 |
+
gauss_latent,
|
529 |
+
self.t_start,
|
530 |
+
encoder_hidden_states=self.prompt_embeds,
|
531 |
+
down_block_additional_residuals=down_block_res_samples,
|
532 |
+
mid_block_additional_residual=mid_block_res_sample,
|
533 |
+
return_dict=False,
|
534 |
+
)[0]
|
535 |
+
|
536 |
+
|
537 |
+
del (
|
538 |
+
image_latent,
|
539 |
+
)
|
540 |
+
|
541 |
+
# decoder
|
542 |
+
prediction = self.decode_prediction(latent_x_t)
|
543 |
+
prediction = self.image_processor.unpad_image(prediction, padding) # [N*E,3,PH,PW]
|
544 |
+
|
545 |
+
prediction = self.image_processor.resize_antialias(
|
546 |
+
prediction, original_resolution, resample_method_output, is_aa=False
|
547 |
+
) # [N,3,H,W]
|
548 |
+
prediction = self.normalize_normals(prediction) # [N,3,H,W]
|
549 |
+
|
550 |
+
if output_type == "np":
|
551 |
+
prediction = self.image_processor.pt_to_numpy(prediction) # [N,H,W,3]
|
552 |
+
|
553 |
+
# 11. Offload all models
|
554 |
+
self.maybe_free_model_hooks()
|
555 |
+
|
556 |
+
return YosoNormalsOutput(
|
557 |
+
prediction=prediction,
|
558 |
+
latent=latent_x_t,
|
559 |
+
gauss_latent=gauss_latent,
|
560 |
+
)
|
561 |
+
|
562 |
+
# Copied from diffusers.pipelines.marigold.pipeline_marigold_depth.MarigoldDepthPipeline.prepare_latents
|
563 |
+
def prepare_latents(
|
564 |
+
self,
|
565 |
+
image: torch.Tensor,
|
566 |
+
latents: Optional[torch.Tensor],
|
567 |
+
generator: Optional[torch.Generator],
|
568 |
+
ensemble_size: int,
|
569 |
+
batch_size: int,
|
570 |
+
) -> Tuple[torch.Tensor, torch.Tensor]:
|
571 |
+
def retrieve_latents(encoder_output):
|
572 |
+
if hasattr(encoder_output, "latent_dist"):
|
573 |
+
return encoder_output.latent_dist.mode()
|
574 |
+
elif hasattr(encoder_output, "latents"):
|
575 |
+
return encoder_output.latents
|
576 |
+
else:
|
577 |
+
raise AttributeError("Could not access latents of provided encoder_output")
|
578 |
+
|
579 |
+
|
580 |
+
|
581 |
+
image_latent = torch.cat(
|
582 |
+
[
|
583 |
+
retrieve_latents(self.vae.encode(image[i : i + batch_size]))
|
584 |
+
for i in range(0, image.shape[0], batch_size)
|
585 |
+
],
|
586 |
+
dim=0,
|
587 |
+
) # [N,4,h,w]
|
588 |
+
image_latent = image_latent * self.vae.config.scaling_factor
|
589 |
+
image_latent = image_latent.repeat_interleave(ensemble_size, dim=0) # [N*E,4,h,w]
|
590 |
+
|
591 |
+
pred_latent = self.gauss_latent
|
592 |
+
if pred_latent is None:
|
593 |
+
self.gauss_latent = torch.randn_like(image_latent)
|
594 |
+
pred_latent = self.gauss_latent
|
595 |
+
|
596 |
+
return image_latent, pred_latent
|
597 |
+
|
598 |
+
def decode_prediction(self, pred_latent: torch.Tensor) -> torch.Tensor:
|
599 |
+
if pred_latent.dim() != 4 or pred_latent.shape[1] != self.vae.config.latent_channels:
|
600 |
+
raise ValueError(
|
601 |
+
f"Expecting 4D tensor of shape [B,{self.vae.config.latent_channels},H,W]; got {pred_latent.shape}."
|
602 |
+
)
|
603 |
+
|
604 |
+
prediction = self.vae.decode(pred_latent / self.vae.config.scaling_factor, return_dict=False)[0] # [B,3,H,W]
|
605 |
+
|
606 |
+
prediction = self.normalize_normals(prediction) # [B,3,H,W]
|
607 |
+
|
608 |
+
return prediction # [B,3,H,W]
|
609 |
+
|
610 |
+
@staticmethod
|
611 |
+
def normalize_normals(normals: torch.Tensor, eps: float = 1e-6) -> torch.Tensor:
|
612 |
+
if normals.dim() != 4 or normals.shape[1] != 3:
|
613 |
+
raise ValueError(f"Expecting 4D tensor of shape [B,3,H,W]; got {normals.shape}.")
|
614 |
+
|
615 |
+
norm = torch.norm(normals, dim=1, keepdim=True)
|
616 |
+
normals /= norm.clamp(min=eps)
|
617 |
+
|
618 |
+
return normals
|
619 |
+
|
620 |
+
@staticmethod
|
621 |
+
def ensemble_normals(
|
622 |
+
normals: torch.Tensor, output_uncertainty: bool, reduction: str = "closest"
|
623 |
+
) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
|
624 |
+
"""
|
625 |
+
Ensembles the normals maps represented by the `normals` tensor with expected shape `(B, 3, H, W)`, where B is
|
626 |
+
the number of ensemble members for a given prediction of size `(H x W)`.
|
627 |
+
|
628 |
+
Args:
|
629 |
+
normals (`torch.Tensor`):
|
630 |
+
Input ensemble normals maps.
|
631 |
+
output_uncertainty (`bool`, *optional*, defaults to `False`):
|
632 |
+
Whether to output uncertainty map.
|
633 |
+
reduction (`str`, *optional*, defaults to `"closest"`):
|
634 |
+
Reduction method used to ensemble aligned predictions. The accepted values are: `"closest"` and
|
635 |
+
`"mean"`.
|
636 |
+
|
637 |
+
Returns:
|
638 |
+
A tensor of aligned and ensembled normals maps with shape `(1, 3, H, W)` and optionally a tensor of
|
639 |
+
uncertainties of shape `(1, 1, H, W)`.
|
640 |
+
"""
|
641 |
+
if normals.dim() != 4 or normals.shape[1] != 3:
|
642 |
+
raise ValueError(f"Expecting 4D tensor of shape [B,3,H,W]; got {normals.shape}.")
|
643 |
+
if reduction not in ("closest", "mean"):
|
644 |
+
raise ValueError(f"Unrecognized reduction method: {reduction}.")
|
645 |
+
|
646 |
+
mean_normals = normals.mean(dim=0, keepdim=True) # [1,3,H,W]
|
647 |
+
mean_normals = MarigoldNormalsPipeline.normalize_normals(mean_normals) # [1,3,H,W]
|
648 |
+
|
649 |
+
sim_cos = (mean_normals * normals).sum(dim=1, keepdim=True) # [E,1,H,W]
|
650 |
+
sim_cos = sim_cos.clamp(-1, 1) # required to avoid NaN in uncertainty with fp16
|
651 |
+
|
652 |
+
uncertainty = None
|
653 |
+
if output_uncertainty:
|
654 |
+
uncertainty = sim_cos.arccos() # [E,1,H,W]
|
655 |
+
uncertainty = uncertainty.mean(dim=0, keepdim=True) / np.pi # [1,1,H,W]
|
656 |
+
|
657 |
+
if reduction == "mean":
|
658 |
+
return mean_normals, uncertainty # [1,3,H,W], [1,1,H,W]
|
659 |
+
|
660 |
+
closest_indices = sim_cos.argmax(dim=0, keepdim=True) # [1,1,H,W]
|
661 |
+
closest_indices = closest_indices.repeat(1, 3, 1, 1) # [1,3,H,W]
|
662 |
+
closest_normals = torch.gather(normals, 0, closest_indices) # [1,3,H,W]
|
663 |
+
|
664 |
+
return closest_normals, uncertainty # [1,3,H,W], [1,1,H,W]
|
665 |
+
|
666 |
+
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
|
667 |
+
def retrieve_timesteps(
|
668 |
+
scheduler,
|
669 |
+
num_inference_steps: Optional[int] = None,
|
670 |
+
device: Optional[Union[str, torch.device]] = None,
|
671 |
+
timesteps: Optional[List[int]] = None,
|
672 |
+
sigmas: Optional[List[float]] = None,
|
673 |
+
**kwargs,
|
674 |
+
):
|
675 |
+
"""
|
676 |
+
Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
|
677 |
+
custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
|
678 |
+
|
679 |
+
Args:
|
680 |
+
scheduler (`SchedulerMixin`):
|
681 |
+
The scheduler to get timesteps from.
|
682 |
+
num_inference_steps (`int`):
|
683 |
+
The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
|
684 |
+
must be `None`.
|
685 |
+
device (`str` or `torch.device`, *optional*):
|
686 |
+
The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
|
687 |
+
timesteps (`List[int]`, *optional*):
|
688 |
+
Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
|
689 |
+
`num_inference_steps` and `sigmas` must be `None`.
|
690 |
+
sigmas (`List[float]`, *optional*):
|
691 |
+
Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
|
692 |
+
`num_inference_steps` and `timesteps` must be `None`.
|
693 |
+
|
694 |
+
Returns:
|
695 |
+
`Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
|
696 |
+
second element is the number of inference steps.
|
697 |
+
"""
|
698 |
+
if timesteps is not None and sigmas is not None:
|
699 |
+
raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
|
700 |
+
if timesteps is not None:
|
701 |
+
accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
|
702 |
+
if not accepts_timesteps:
|
703 |
+
raise ValueError(
|
704 |
+
f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
|
705 |
+
f" timestep schedules. Please check whether you are using the correct scheduler."
|
706 |
+
)
|
707 |
+
scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
|
708 |
+
timesteps = scheduler.timesteps
|
709 |
+
num_inference_steps = len(timesteps)
|
710 |
+
elif sigmas is not None:
|
711 |
+
accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
|
712 |
+
if not accept_sigmas:
|
713 |
+
raise ValueError(
|
714 |
+
f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
|
715 |
+
f" sigmas schedules. Please check whether you are using the correct scheduler."
|
716 |
+
)
|
717 |
+
scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
|
718 |
+
timesteps = scheduler.timesteps
|
719 |
+
num_inference_steps = len(timesteps)
|
720 |
+
else:
|
721 |
+
scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
|
722 |
+
timesteps = scheduler.timesteps
|
723 |
+
return timesteps, num_inference_steps
|
stablenormal/scheduler/__init__.py
ADDED
File without changes
|
stablenormal/stablecontrolnet.py
ADDED
@@ -0,0 +1,1354 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Copyright 2024 The HuggingFace Team. All rights reserved.
|
2 |
+
#
|
3 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
4 |
+
# you may not use this file except in compliance with the License.
|
5 |
+
# You may obtain a copy of the License at
|
6 |
+
#
|
7 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
8 |
+
#
|
9 |
+
# Unless required by applicable law or agreed to in writing, software
|
10 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
11 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
12 |
+
# See the License for the specific language governing permissions and
|
13 |
+
# limitations under the License.
|
14 |
+
|
15 |
+
|
16 |
+
import inspect
|
17 |
+
from typing import Any, Callable, Dict, List, Optional, Tuple, Union
|
18 |
+
|
19 |
+
import numpy as np
|
20 |
+
import PIL.Image
|
21 |
+
import torch
|
22 |
+
import torch.nn.functional as F
|
23 |
+
from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
|
24 |
+
|
25 |
+
from ...callbacks import MultiPipelineCallbacks, PipelineCallback
|
26 |
+
from ...image_processor import PipelineImageInput, VaeImageProcessor
|
27 |
+
from ...loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
|
28 |
+
from ...models import AutoencoderKL, ControlNetModel, ImageProjection, UNet2DConditionModel
|
29 |
+
from ...models.lora import adjust_lora_scale_text_encoder
|
30 |
+
from ...schedulers import KarrasDiffusionSchedulers
|
31 |
+
from ...utils import (
|
32 |
+
USE_PEFT_BACKEND,
|
33 |
+
deprecate,
|
34 |
+
logging,
|
35 |
+
replace_example_docstring,
|
36 |
+
scale_lora_layers,
|
37 |
+
unscale_lora_layers,
|
38 |
+
)
|
39 |
+
from ...utils.torch_utils import is_compiled_module, is_torch_version, randn_tensor
|
40 |
+
from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
|
41 |
+
from ..stable_diffusion.pipeline_output import StableDiffusionPipelineOutput
|
42 |
+
from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
|
43 |
+
from .multicontrolnet import MultiControlNetModel
|
44 |
+
|
45 |
+
|
46 |
+
logger = logging.get_logger(__name__) # pylint: disable=invalid-name
|
47 |
+
|
48 |
+
|
49 |
+
EXAMPLE_DOC_STRING = """
|
50 |
+
Examples:
|
51 |
+
```py
|
52 |
+
>>> # !pip install opencv-python transformers accelerate
|
53 |
+
>>> from diffusers import StableDiffusionControlNetPipeline, ControlNetModel, UniPCMultistepScheduler
|
54 |
+
>>> from diffusers.utils import load_image
|
55 |
+
>>> import numpy as np
|
56 |
+
>>> import torch
|
57 |
+
|
58 |
+
>>> import cv2
|
59 |
+
>>> from PIL import Image
|
60 |
+
|
61 |
+
>>> # download an image
|
62 |
+
>>> image = load_image(
|
63 |
+
... "https://hf.co/datasets/huggingface/documentation-images/resolve/main/diffusers/input_image_vermeer.png"
|
64 |
+
... )
|
65 |
+
>>> image = np.array(image)
|
66 |
+
|
67 |
+
>>> # get canny image
|
68 |
+
>>> image = cv2.Canny(image, 100, 200)
|
69 |
+
>>> image = image[:, :, None]
|
70 |
+
>>> image = np.concatenate([image, image, image], axis=2)
|
71 |
+
>>> canny_image = Image.fromarray(image)
|
72 |
+
|
73 |
+
>>> # load control net and stable diffusion v1-5
|
74 |
+
>>> controlnet = ControlNetModel.from_pretrained("lllyasviel/sd-controlnet-canny", torch_dtype=torch.float16)
|
75 |
+
>>> pipe = StableDiffusionControlNetPipeline.from_pretrained(
|
76 |
+
... "runwayml/stable-diffusion-v1-5", controlnet=controlnet, torch_dtype=torch.float16
|
77 |
+
... )
|
78 |
+
|
79 |
+
>>> # speed up diffusion process with faster scheduler and memory optimization
|
80 |
+
>>> pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config)
|
81 |
+
>>> # remove following line if xformers is not installed
|
82 |
+
>>> pipe.enable_xformers_memory_efficient_attention()
|
83 |
+
|
84 |
+
>>> pipe.enable_model_cpu_offload()
|
85 |
+
|
86 |
+
>>> # generate image
|
87 |
+
>>> generator = torch.manual_seed(0)
|
88 |
+
>>> image = pipe(
|
89 |
+
... "futuristic-looking woman", num_inference_steps=20, generator=generator, image=canny_image
|
90 |
+
... ).images[0]
|
91 |
+
```
|
92 |
+
"""
|
93 |
+
|
94 |
+
|
95 |
+
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
|
96 |
+
def retrieve_timesteps(
|
97 |
+
scheduler,
|
98 |
+
num_inference_steps: Optional[int] = None,
|
99 |
+
device: Optional[Union[str, torch.device]] = None,
|
100 |
+
timesteps: Optional[List[int]] = None,
|
101 |
+
sigmas: Optional[List[float]] = None,
|
102 |
+
**kwargs,
|
103 |
+
):
|
104 |
+
"""
|
105 |
+
Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
|
106 |
+
custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
|
107 |
+
|
108 |
+
Args:
|
109 |
+
scheduler (`SchedulerMixin`):
|
110 |
+
The scheduler to get timesteps from.
|
111 |
+
num_inference_steps (`int`):
|
112 |
+
The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
|
113 |
+
must be `None`.
|
114 |
+
device (`str` or `torch.device`, *optional*):
|
115 |
+
The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
|
116 |
+
timesteps (`List[int]`, *optional*):
|
117 |
+
Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
|
118 |
+
`num_inference_steps` and `sigmas` must be `None`.
|
119 |
+
sigmas (`List[float]`, *optional*):
|
120 |
+
Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
|
121 |
+
`num_inference_steps` and `timesteps` must be `None`.
|
122 |
+
|
123 |
+
Returns:
|
124 |
+
`Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
|
125 |
+
second element is the number of inference steps.
|
126 |
+
"""
|
127 |
+
if timesteps is not None and sigmas is not None:
|
128 |
+
raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
|
129 |
+
if timesteps is not None:
|
130 |
+
accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
|
131 |
+
if not accepts_timesteps:
|
132 |
+
raise ValueError(
|
133 |
+
f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
|
134 |
+
f" timestep schedules. Please check whether you are using the correct scheduler."
|
135 |
+
)
|
136 |
+
scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
|
137 |
+
timesteps = scheduler.timesteps
|
138 |
+
num_inference_steps = len(timesteps)
|
139 |
+
elif sigmas is not None:
|
140 |
+
accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
|
141 |
+
if not accept_sigmas:
|
142 |
+
raise ValueError(
|
143 |
+
f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
|
144 |
+
f" sigmas schedules. Please check whether you are using the correct scheduler."
|
145 |
+
)
|
146 |
+
scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
|
147 |
+
timesteps = scheduler.timesteps
|
148 |
+
num_inference_steps = len(timesteps)
|
149 |
+
else:
|
150 |
+
scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
|
151 |
+
timesteps = scheduler.timesteps
|
152 |
+
return timesteps, num_inference_steps
|
153 |
+
|
154 |
+
|
155 |
+
class StableDiffusionControlNetPipeline(
|
156 |
+
DiffusionPipeline,
|
157 |
+
StableDiffusionMixin,
|
158 |
+
TextualInversionLoaderMixin,
|
159 |
+
LoraLoaderMixin,
|
160 |
+
IPAdapterMixin,
|
161 |
+
FromSingleFileMixin,
|
162 |
+
):
|
163 |
+
r"""
|
164 |
+
Pipeline for text-to-image generation using Stable Diffusion with ControlNet guidance.
|
165 |
+
|
166 |
+
This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
|
167 |
+
implemented for all pipelines (downloading, saving, running on a particular device, etc.).
|
168 |
+
|
169 |
+
The pipeline also inherits the following loading methods:
|
170 |
+
- [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
|
171 |
+
- [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
|
172 |
+
- [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
|
173 |
+
- [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
|
174 |
+
- [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
|
175 |
+
|
176 |
+
Args:
|
177 |
+
vae ([`AutoencoderKL`]):
|
178 |
+
Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations.
|
179 |
+
text_encoder ([`~transformers.CLIPTextModel`]):
|
180 |
+
Frozen text-encoder ([clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14)).
|
181 |
+
tokenizer ([`~transformers.CLIPTokenizer`]):
|
182 |
+
A `CLIPTokenizer` to tokenize text.
|
183 |
+
unet ([`UNet2DConditionModel`]):
|
184 |
+
A `UNet2DConditionModel` to denoise the encoded image latents.
|
185 |
+
controlnet ([`ControlNetModel`] or `List[ControlNetModel]`):
|
186 |
+
Provides additional conditioning to the `unet` during the denoising process. If you set multiple
|
187 |
+
ControlNets as a list, the outputs from each ControlNet are added together to create one combined
|
188 |
+
additional conditioning.
|
189 |
+
scheduler ([`SchedulerMixin`]):
|
190 |
+
A scheduler to be used in combination with `unet` to denoise the encoded image latents. Can be one of
|
191 |
+
[`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
|
192 |
+
safety_checker ([`StableDiffusionSafetyChecker`]):
|
193 |
+
Classification module that estimates whether generated images could be considered offensive or harmful.
|
194 |
+
Please refer to the [model card](https://huggingface.co/runwayml/stable-diffusion-v1-5) for more details
|
195 |
+
about a model's potential harms.
|
196 |
+
feature_extractor ([`~transformers.CLIPImageProcessor`]):
|
197 |
+
A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
|
198 |
+
"""
|
199 |
+
|
200 |
+
model_cpu_offload_seq = "text_encoder->image_encoder->unet->vae"
|
201 |
+
_optional_components = ["safety_checker", "feature_extractor", "image_encoder"]
|
202 |
+
_exclude_from_cpu_offload = ["safety_checker"]
|
203 |
+
_callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"]
|
204 |
+
|
205 |
+
def __init__(
|
206 |
+
self,
|
207 |
+
vae: AutoencoderKL,
|
208 |
+
text_encoder: CLIPTextModel,
|
209 |
+
tokenizer: CLIPTokenizer,
|
210 |
+
unet: UNet2DConditionModel,
|
211 |
+
controlnet: Union[ControlNetModel, List[ControlNetModel], Tuple[ControlNetModel], MultiControlNetModel],
|
212 |
+
scheduler: KarrasDiffusionSchedulers,
|
213 |
+
safety_checker: StableDiffusionSafetyChecker,
|
214 |
+
feature_extractor: CLIPImageProcessor,
|
215 |
+
image_encoder: CLIPVisionModelWithProjection = None,
|
216 |
+
requires_safety_checker: bool = True,
|
217 |
+
):
|
218 |
+
super().__init__()
|
219 |
+
|
220 |
+
if safety_checker is None and requires_safety_checker:
|
221 |
+
logger.warning(
|
222 |
+
f"You have disabled the safety checker for {self.__class__} by passing `safety_checker=None`. Ensure"
|
223 |
+
" that you abide to the conditions of the Stable Diffusion license and do not expose unfiltered"
|
224 |
+
" results in services or applications open to the public. Both the diffusers team and Hugging Face"
|
225 |
+
" strongly recommend to keep the safety filter enabled in all public facing circumstances, disabling"
|
226 |
+
" it only for use-cases that involve analyzing network behavior or auditing its results. For more"
|
227 |
+
" information, please have a look at https://github.com/huggingface/diffusers/pull/254 ."
|
228 |
+
)
|
229 |
+
|
230 |
+
if safety_checker is not None and feature_extractor is None:
|
231 |
+
raise ValueError(
|
232 |
+
"Make sure to define a feature extractor when loading {self.__class__} if you want to use the safety"
|
233 |
+
" checker. If you do not want to use the safety checker, you can pass `'safety_checker=None'` instead."
|
234 |
+
)
|
235 |
+
|
236 |
+
if isinstance(controlnet, (list, tuple)):
|
237 |
+
controlnet = MultiControlNetModel(controlnet)
|
238 |
+
|
239 |
+
self.register_modules(
|
240 |
+
vae=vae,
|
241 |
+
text_encoder=text_encoder,
|
242 |
+
tokenizer=tokenizer,
|
243 |
+
unet=unet,
|
244 |
+
controlnet=controlnet,
|
245 |
+
scheduler=scheduler,
|
246 |
+
safety_checker=safety_checker,
|
247 |
+
feature_extractor=feature_extractor,
|
248 |
+
image_encoder=image_encoder,
|
249 |
+
)
|
250 |
+
self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
|
251 |
+
self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True)
|
252 |
+
self.control_image_processor = VaeImageProcessor(
|
253 |
+
vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True, do_normalize=False
|
254 |
+
)
|
255 |
+
self.register_to_config(requires_safety_checker=requires_safety_checker)
|
256 |
+
|
257 |
+
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
|
258 |
+
def _encode_prompt(
|
259 |
+
self,
|
260 |
+
prompt,
|
261 |
+
device,
|
262 |
+
num_images_per_prompt,
|
263 |
+
do_classifier_free_guidance,
|
264 |
+
negative_prompt=None,
|
265 |
+
prompt_embeds: Optional[torch.Tensor] = None,
|
266 |
+
negative_prompt_embeds: Optional[torch.Tensor] = None,
|
267 |
+
lora_scale: Optional[float] = None,
|
268 |
+
**kwargs,
|
269 |
+
):
|
270 |
+
deprecation_message = "`_encode_prompt()` is deprecated and it will be removed in a future version. Use `encode_prompt()` instead. Also, be aware that the output format changed from a concatenated tensor to a tuple."
|
271 |
+
deprecate("_encode_prompt()", "1.0.0", deprecation_message, standard_warn=False)
|
272 |
+
|
273 |
+
prompt_embeds_tuple = self.encode_prompt(
|
274 |
+
prompt=prompt,
|
275 |
+
device=device,
|
276 |
+
num_images_per_prompt=num_images_per_prompt,
|
277 |
+
do_classifier_free_guidance=do_classifier_free_guidance,
|
278 |
+
negative_prompt=negative_prompt,
|
279 |
+
prompt_embeds=prompt_embeds,
|
280 |
+
negative_prompt_embeds=negative_prompt_embeds,
|
281 |
+
lora_scale=lora_scale,
|
282 |
+
**kwargs,
|
283 |
+
)
|
284 |
+
|
285 |
+
# concatenate for backwards comp
|
286 |
+
prompt_embeds = torch.cat([prompt_embeds_tuple[1], prompt_embeds_tuple[0]])
|
287 |
+
|
288 |
+
return prompt_embeds
|
289 |
+
|
290 |
+
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
|
291 |
+
def encode_prompt(
|
292 |
+
self,
|
293 |
+
prompt,
|
294 |
+
device,
|
295 |
+
num_images_per_prompt,
|
296 |
+
do_classifier_free_guidance,
|
297 |
+
negative_prompt=None,
|
298 |
+
prompt_embeds: Optional[torch.Tensor] = None,
|
299 |
+
negative_prompt_embeds: Optional[torch.Tensor] = None,
|
300 |
+
lora_scale: Optional[float] = None,
|
301 |
+
clip_skip: Optional[int] = None,
|
302 |
+
):
|
303 |
+
r"""
|
304 |
+
Encodes the prompt into text encoder hidden states.
|
305 |
+
|
306 |
+
Args:
|
307 |
+
prompt (`str` or `List[str]`, *optional*):
|
308 |
+
prompt to be encoded
|
309 |
+
device: (`torch.device`):
|
310 |
+
torch device
|
311 |
+
num_images_per_prompt (`int`):
|
312 |
+
number of images that should be generated per prompt
|
313 |
+
do_classifier_free_guidance (`bool`):
|
314 |
+
whether to use classifier free guidance or not
|
315 |
+
negative_prompt (`str` or `List[str]`, *optional*):
|
316 |
+
The prompt or prompts not to guide the image generation. If not defined, one has to pass
|
317 |
+
`negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
|
318 |
+
less than `1`).
|
319 |
+
prompt_embeds (`torch.Tensor`, *optional*):
|
320 |
+
Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
|
321 |
+
provided, text embeddings will be generated from `prompt` input argument.
|
322 |
+
negative_prompt_embeds (`torch.Tensor`, *optional*):
|
323 |
+
Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
|
324 |
+
weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
|
325 |
+
argument.
|
326 |
+
lora_scale (`float`, *optional*):
|
327 |
+
A LoRA scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
|
328 |
+
clip_skip (`int`, *optional*):
|
329 |
+
Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
|
330 |
+
the output of the pre-final layer will be used for computing the prompt embeddings.
|
331 |
+
"""
|
332 |
+
# set lora scale so that monkey patched LoRA
|
333 |
+
# function of text encoder can correctly access it
|
334 |
+
if lora_scale is not None and isinstance(self, LoraLoaderMixin):
|
335 |
+
self._lora_scale = lora_scale
|
336 |
+
|
337 |
+
# dynamically adjust the LoRA scale
|
338 |
+
if not USE_PEFT_BACKEND:
|
339 |
+
adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)
|
340 |
+
else:
|
341 |
+
scale_lora_layers(self.text_encoder, lora_scale)
|
342 |
+
|
343 |
+
if prompt is not None and isinstance(prompt, str):
|
344 |
+
batch_size = 1
|
345 |
+
elif prompt is not None and isinstance(prompt, list):
|
346 |
+
batch_size = len(prompt)
|
347 |
+
else:
|
348 |
+
batch_size = prompt_embeds.shape[0]
|
349 |
+
|
350 |
+
if prompt_embeds is None:
|
351 |
+
# textual inversion: process multi-vector tokens if necessary
|
352 |
+
if isinstance(self, TextualInversionLoaderMixin):
|
353 |
+
prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
|
354 |
+
|
355 |
+
text_inputs = self.tokenizer(
|
356 |
+
prompt,
|
357 |
+
padding="max_length",
|
358 |
+
max_length=self.tokenizer.model_max_length,
|
359 |
+
truncation=True,
|
360 |
+
return_tensors="pt",
|
361 |
+
)
|
362 |
+
text_input_ids = text_inputs.input_ids
|
363 |
+
untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
|
364 |
+
|
365 |
+
if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
|
366 |
+
text_input_ids, untruncated_ids
|
367 |
+
):
|
368 |
+
removed_text = self.tokenizer.batch_decode(
|
369 |
+
untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1]
|
370 |
+
)
|
371 |
+
logger.warning(
|
372 |
+
"The following part of your input was truncated because CLIP can only handle sequences up to"
|
373 |
+
f" {self.tokenizer.model_max_length} tokens: {removed_text}"
|
374 |
+
)
|
375 |
+
|
376 |
+
if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
|
377 |
+
attention_mask = text_inputs.attention_mask.to(device)
|
378 |
+
else:
|
379 |
+
attention_mask = None
|
380 |
+
|
381 |
+
if clip_skip is None:
|
382 |
+
prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)
|
383 |
+
prompt_embeds = prompt_embeds[0]
|
384 |
+
else:
|
385 |
+
prompt_embeds = self.text_encoder(
|
386 |
+
text_input_ids.to(device), attention_mask=attention_mask, output_hidden_states=True
|
387 |
+
)
|
388 |
+
# Access the `hidden_states` first, that contains a tuple of
|
389 |
+
# all the hidden states from the encoder layers. Then index into
|
390 |
+
# the tuple to access the hidden states from the desired layer.
|
391 |
+
prompt_embeds = prompt_embeds[-1][-(clip_skip + 1)]
|
392 |
+
# We also need to apply the final LayerNorm here to not mess with the
|
393 |
+
# representations. The `last_hidden_states` that we typically use for
|
394 |
+
# obtaining the final prompt representations passes through the LayerNorm
|
395 |
+
# layer.
|
396 |
+
prompt_embeds = self.text_encoder.text_model.final_layer_norm(prompt_embeds)
|
397 |
+
|
398 |
+
if self.text_encoder is not None:
|
399 |
+
prompt_embeds_dtype = self.text_encoder.dtype
|
400 |
+
elif self.unet is not None:
|
401 |
+
prompt_embeds_dtype = self.unet.dtype
|
402 |
+
else:
|
403 |
+
prompt_embeds_dtype = prompt_embeds.dtype
|
404 |
+
|
405 |
+
prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
|
406 |
+
|
407 |
+
bs_embed, seq_len, _ = prompt_embeds.shape
|
408 |
+
# duplicate text embeddings for each generation per prompt, using mps friendly method
|
409 |
+
prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
|
410 |
+
prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
|
411 |
+
|
412 |
+
# get unconditional embeddings for classifier free guidance
|
413 |
+
if do_classifier_free_guidance and negative_prompt_embeds is None:
|
414 |
+
uncond_tokens: List[str]
|
415 |
+
if negative_prompt is None:
|
416 |
+
uncond_tokens = [""] * batch_size
|
417 |
+
elif prompt is not None and type(prompt) is not type(negative_prompt):
|
418 |
+
raise TypeError(
|
419 |
+
f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
|
420 |
+
f" {type(prompt)}."
|
421 |
+
)
|
422 |
+
elif isinstance(negative_prompt, str):
|
423 |
+
uncond_tokens = [negative_prompt]
|
424 |
+
elif batch_size != len(negative_prompt):
|
425 |
+
raise ValueError(
|
426 |
+
f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
|
427 |
+
f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
|
428 |
+
" the batch size of `prompt`."
|
429 |
+
)
|
430 |
+
else:
|
431 |
+
uncond_tokens = negative_prompt
|
432 |
+
|
433 |
+
# textual inversion: process multi-vector tokens if necessary
|
434 |
+
if isinstance(self, TextualInversionLoaderMixin):
|
435 |
+
uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
|
436 |
+
|
437 |
+
max_length = prompt_embeds.shape[1]
|
438 |
+
uncond_input = self.tokenizer(
|
439 |
+
uncond_tokens,
|
440 |
+
padding="max_length",
|
441 |
+
max_length=max_length,
|
442 |
+
truncation=True,
|
443 |
+
return_tensors="pt",
|
444 |
+
)
|
445 |
+
|
446 |
+
if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
|
447 |
+
attention_mask = uncond_input.attention_mask.to(device)
|
448 |
+
else:
|
449 |
+
attention_mask = None
|
450 |
+
|
451 |
+
negative_prompt_embeds = self.text_encoder(
|
452 |
+
uncond_input.input_ids.to(device),
|
453 |
+
attention_mask=attention_mask,
|
454 |
+
)
|
455 |
+
negative_prompt_embeds = negative_prompt_embeds[0]
|
456 |
+
|
457 |
+
if do_classifier_free_guidance:
|
458 |
+
# duplicate unconditional embeddings for each generation per prompt, using mps friendly method
|
459 |
+
seq_len = negative_prompt_embeds.shape[1]
|
460 |
+
|
461 |
+
negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
|
462 |
+
|
463 |
+
negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
|
464 |
+
negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
|
465 |
+
|
466 |
+
if self.text_encoder is not None:
|
467 |
+
if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
|
468 |
+
# Retrieve the original scale by scaling back the LoRA layers
|
469 |
+
unscale_lora_layers(self.text_encoder, lora_scale)
|
470 |
+
|
471 |
+
return prompt_embeds, negative_prompt_embeds
|
472 |
+
|
473 |
+
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_image
|
474 |
+
def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):
|
475 |
+
dtype = next(self.image_encoder.parameters()).dtype
|
476 |
+
|
477 |
+
if not isinstance(image, torch.Tensor):
|
478 |
+
image = self.feature_extractor(image, return_tensors="pt").pixel_values
|
479 |
+
|
480 |
+
image = image.to(device=device, dtype=dtype)
|
481 |
+
if output_hidden_states:
|
482 |
+
image_enc_hidden_states = self.image_encoder(image, output_hidden_states=True).hidden_states[-2]
|
483 |
+
image_enc_hidden_states = image_enc_hidden_states.repeat_interleave(num_images_per_prompt, dim=0)
|
484 |
+
uncond_image_enc_hidden_states = self.image_encoder(
|
485 |
+
torch.zeros_like(image), output_hidden_states=True
|
486 |
+
).hidden_states[-2]
|
487 |
+
uncond_image_enc_hidden_states = uncond_image_enc_hidden_states.repeat_interleave(
|
488 |
+
num_images_per_prompt, dim=0
|
489 |
+
)
|
490 |
+
return image_enc_hidden_states, uncond_image_enc_hidden_states
|
491 |
+
else:
|
492 |
+
image_embeds = self.image_encoder(image).image_embeds
|
493 |
+
image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)
|
494 |
+
uncond_image_embeds = torch.zeros_like(image_embeds)
|
495 |
+
|
496 |
+
return image_embeds, uncond_image_embeds
|
497 |
+
|
498 |
+
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
|
499 |
+
def prepare_ip_adapter_image_embeds(
|
500 |
+
self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt, do_classifier_free_guidance
|
501 |
+
):
|
502 |
+
if ip_adapter_image_embeds is None:
|
503 |
+
if not isinstance(ip_adapter_image, list):
|
504 |
+
ip_adapter_image = [ip_adapter_image]
|
505 |
+
|
506 |
+
if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
|
507 |
+
raise ValueError(
|
508 |
+
f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
|
509 |
+
)
|
510 |
+
|
511 |
+
image_embeds = []
|
512 |
+
for single_ip_adapter_image, image_proj_layer in zip(
|
513 |
+
ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
|
514 |
+
):
|
515 |
+
output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
|
516 |
+
single_image_embeds, single_negative_image_embeds = self.encode_image(
|
517 |
+
single_ip_adapter_image, device, 1, output_hidden_state
|
518 |
+
)
|
519 |
+
single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
|
520 |
+
single_negative_image_embeds = torch.stack(
|
521 |
+
[single_negative_image_embeds] * num_images_per_prompt, dim=0
|
522 |
+
)
|
523 |
+
|
524 |
+
if do_classifier_free_guidance:
|
525 |
+
single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
|
526 |
+
single_image_embeds = single_image_embeds.to(device)
|
527 |
+
|
528 |
+
image_embeds.append(single_image_embeds)
|
529 |
+
else:
|
530 |
+
repeat_dims = [1]
|
531 |
+
image_embeds = []
|
532 |
+
for single_image_embeds in ip_adapter_image_embeds:
|
533 |
+
if do_classifier_free_guidance:
|
534 |
+
single_negative_image_embeds, single_image_embeds = single_image_embeds.chunk(2)
|
535 |
+
single_image_embeds = single_image_embeds.repeat(
|
536 |
+
num_images_per_prompt, *(repeat_dims * len(single_image_embeds.shape[1:]))
|
537 |
+
)
|
538 |
+
single_negative_image_embeds = single_negative_image_embeds.repeat(
|
539 |
+
num_images_per_prompt, *(repeat_dims * len(single_negative_image_embeds.shape[1:]))
|
540 |
+
)
|
541 |
+
single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
|
542 |
+
else:
|
543 |
+
single_image_embeds = single_image_embeds.repeat(
|
544 |
+
num_images_per_prompt, *(repeat_dims * len(single_image_embeds.shape[1:]))
|
545 |
+
)
|
546 |
+
image_embeds.append(single_image_embeds)
|
547 |
+
|
548 |
+
return image_embeds
|
549 |
+
|
550 |
+
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
|
551 |
+
def run_safety_checker(self, image, device, dtype):
|
552 |
+
if self.safety_checker is None:
|
553 |
+
has_nsfw_concept = None
|
554 |
+
else:
|
555 |
+
if torch.is_tensor(image):
|
556 |
+
feature_extractor_input = self.image_processor.postprocess(image, output_type="pil")
|
557 |
+
else:
|
558 |
+
feature_extractor_input = self.image_processor.numpy_to_pil(image)
|
559 |
+
safety_checker_input = self.feature_extractor(feature_extractor_input, return_tensors="pt").to(device)
|
560 |
+
image, has_nsfw_concept = self.safety_checker(
|
561 |
+
images=image, clip_input=safety_checker_input.pixel_values.to(dtype)
|
562 |
+
)
|
563 |
+
return image, has_nsfw_concept
|
564 |
+
|
565 |
+
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
|
566 |
+
def decode_latents(self, latents):
|
567 |
+
deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
|
568 |
+
deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
|
569 |
+
|
570 |
+
latents = 1 / self.vae.config.scaling_factor * latents
|
571 |
+
image = self.vae.decode(latents, return_dict=False)[0]
|
572 |
+
image = (image / 2 + 0.5).clamp(0, 1)
|
573 |
+
# we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
|
574 |
+
image = image.cpu().permute(0, 2, 3, 1).float().numpy()
|
575 |
+
return image
|
576 |
+
|
577 |
+
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
|
578 |
+
def prepare_extra_step_kwargs(self, generator, eta):
|
579 |
+
# prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
|
580 |
+
# eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
|
581 |
+
# eta corresponds to η in DDIM paper: https://arxiv.org/abs/2010.02502
|
582 |
+
# and should be between [0, 1]
|
583 |
+
|
584 |
+
accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
|
585 |
+
extra_step_kwargs = {}
|
586 |
+
if accepts_eta:
|
587 |
+
extra_step_kwargs["eta"] = eta
|
588 |
+
|
589 |
+
# check if the scheduler accepts generator
|
590 |
+
accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
|
591 |
+
if accepts_generator:
|
592 |
+
extra_step_kwargs["generator"] = generator
|
593 |
+
return extra_step_kwargs
|
594 |
+
|
595 |
+
def check_inputs(
|
596 |
+
self,
|
597 |
+
prompt,
|
598 |
+
image,
|
599 |
+
callback_steps,
|
600 |
+
negative_prompt=None,
|
601 |
+
prompt_embeds=None,
|
602 |
+
negative_prompt_embeds=None,
|
603 |
+
ip_adapter_image=None,
|
604 |
+
ip_adapter_image_embeds=None,
|
605 |
+
controlnet_conditioning_scale=1.0,
|
606 |
+
control_guidance_start=0.0,
|
607 |
+
control_guidance_end=1.0,
|
608 |
+
callback_on_step_end_tensor_inputs=None,
|
609 |
+
):
|
610 |
+
if callback_steps is not None and (not isinstance(callback_steps, int) or callback_steps <= 0):
|
611 |
+
raise ValueError(
|
612 |
+
f"`callback_steps` has to be a positive integer but is {callback_steps} of type"
|
613 |
+
f" {type(callback_steps)}."
|
614 |
+
)
|
615 |
+
|
616 |
+
if callback_on_step_end_tensor_inputs is not None and not all(
|
617 |
+
k in self._callback_tensor_inputs for k in callback_on_step_end_tensor_inputs
|
618 |
+
):
|
619 |
+
raise ValueError(
|
620 |
+
f"`callback_on_step_end_tensor_inputs` has to be in {self._callback_tensor_inputs}, but found {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}"
|
621 |
+
)
|
622 |
+
|
623 |
+
if prompt is not None and prompt_embeds is not None:
|
624 |
+
raise ValueError(
|
625 |
+
f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
|
626 |
+
" only forward one of the two."
|
627 |
+
)
|
628 |
+
elif prompt is None and prompt_embeds is None:
|
629 |
+
raise ValueError(
|
630 |
+
"Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined."
|
631 |
+
)
|
632 |
+
elif prompt is not None and (not isinstance(prompt, str) and not isinstance(prompt, list)):
|
633 |
+
raise ValueError(f"`prompt` has to be of type `str` or `list` but is {type(prompt)}")
|
634 |
+
|
635 |
+
if negative_prompt is not None and negative_prompt_embeds is not None:
|
636 |
+
raise ValueError(
|
637 |
+
f"Cannot forward both `negative_prompt`: {negative_prompt} and `negative_prompt_embeds`:"
|
638 |
+
f" {negative_prompt_embeds}. Please make sure to only forward one of the two."
|
639 |
+
)
|
640 |
+
|
641 |
+
if prompt_embeds is not None and negative_prompt_embeds is not None:
|
642 |
+
if prompt_embeds.shape != negative_prompt_embeds.shape:
|
643 |
+
raise ValueError(
|
644 |
+
"`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly, but"
|
645 |
+
f" got: `prompt_embeds` {prompt_embeds.shape} != `negative_prompt_embeds`"
|
646 |
+
f" {negative_prompt_embeds.shape}."
|
647 |
+
)
|
648 |
+
|
649 |
+
# Check `image`
|
650 |
+
is_compiled = hasattr(F, "scaled_dot_product_attention") and isinstance(
|
651 |
+
self.controlnet, torch._dynamo.eval_frame.OptimizedModule
|
652 |
+
)
|
653 |
+
if (
|
654 |
+
isinstance(self.controlnet, ControlNetModel)
|
655 |
+
or is_compiled
|
656 |
+
and isinstance(self.controlnet._orig_mod, ControlNetModel)
|
657 |
+
):
|
658 |
+
self.check_image(image, prompt, prompt_embeds)
|
659 |
+
elif (
|
660 |
+
isinstance(self.controlnet, MultiControlNetModel)
|
661 |
+
or is_compiled
|
662 |
+
and isinstance(self.controlnet._orig_mod, MultiControlNetModel)
|
663 |
+
):
|
664 |
+
if not isinstance(image, list):
|
665 |
+
raise TypeError("For multiple controlnets: `image` must be type `list`")
|
666 |
+
|
667 |
+
# When `image` is a nested list:
|
668 |
+
# (e.g. [[canny_image_1, pose_image_1], [canny_image_2, pose_image_2]])
|
669 |
+
elif any(isinstance(i, list) for i in image):
|
670 |
+
transposed_image = [list(t) for t in zip(*image)]
|
671 |
+
if len(transposed_image) != len(self.controlnet.nets):
|
672 |
+
raise ValueError(
|
673 |
+
f"For multiple controlnets: if you pass`image` as a list of list, each sublist must have the same length as the number of controlnets, but the sublists in `image` got {len(transposed_image)} images and {len(self.controlnet.nets)} ControlNets."
|
674 |
+
)
|
675 |
+
for image_ in transposed_image:
|
676 |
+
self.check_image(image_, prompt, prompt_embeds)
|
677 |
+
elif len(image) != len(self.controlnet.nets):
|
678 |
+
raise ValueError(
|
679 |
+
f"For multiple controlnets: `image` must have the same length as the number of controlnets, but got {len(image)} images and {len(self.controlnet.nets)} ControlNets."
|
680 |
+
)
|
681 |
+
else:
|
682 |
+
for image_ in image:
|
683 |
+
self.check_image(image_, prompt, prompt_embeds)
|
684 |
+
else:
|
685 |
+
assert False
|
686 |
+
|
687 |
+
# Check `controlnet_conditioning_scale`
|
688 |
+
if (
|
689 |
+
isinstance(self.controlnet, ControlNetModel)
|
690 |
+
or is_compiled
|
691 |
+
and isinstance(self.controlnet._orig_mod, ControlNetModel)
|
692 |
+
):
|
693 |
+
if not isinstance(controlnet_conditioning_scale, float):
|
694 |
+
raise TypeError("For single controlnet: `controlnet_conditioning_scale` must be type `float`.")
|
695 |
+
elif (
|
696 |
+
isinstance(self.controlnet, MultiControlNetModel)
|
697 |
+
or is_compiled
|
698 |
+
and isinstance(self.controlnet._orig_mod, MultiControlNetModel)
|
699 |
+
):
|
700 |
+
if isinstance(controlnet_conditioning_scale, list):
|
701 |
+
if any(isinstance(i, list) for i in controlnet_conditioning_scale):
|
702 |
+
raise ValueError(
|
703 |
+
"A single batch of varying conditioning scale settings (e.g. [[1.0, 0.5], [0.2, 0.8]]) is not supported at the moment. "
|
704 |
+
"The conditioning scale must be fixed across the batch."
|
705 |
+
)
|
706 |
+
elif isinstance(controlnet_conditioning_scale, list) and len(controlnet_conditioning_scale) != len(
|
707 |
+
self.controlnet.nets
|
708 |
+
):
|
709 |
+
raise ValueError(
|
710 |
+
"For multiple controlnets: When `controlnet_conditioning_scale` is specified as `list`, it must have"
|
711 |
+
" the same length as the number of controlnets"
|
712 |
+
)
|
713 |
+
else:
|
714 |
+
assert False
|
715 |
+
|
716 |
+
if not isinstance(control_guidance_start, (tuple, list)):
|
717 |
+
control_guidance_start = [control_guidance_start]
|
718 |
+
|
719 |
+
if not isinstance(control_guidance_end, (tuple, list)):
|
720 |
+
control_guidance_end = [control_guidance_end]
|
721 |
+
|
722 |
+
if len(control_guidance_start) != len(control_guidance_end):
|
723 |
+
raise ValueError(
|
724 |
+
f"`control_guidance_start` has {len(control_guidance_start)} elements, but `control_guidance_end` has {len(control_guidance_end)} elements. Make sure to provide the same number of elements to each list."
|
725 |
+
)
|
726 |
+
|
727 |
+
if isinstance(self.controlnet, MultiControlNetModel):
|
728 |
+
if len(control_guidance_start) != len(self.controlnet.nets):
|
729 |
+
raise ValueError(
|
730 |
+
f"`control_guidance_start`: {control_guidance_start} has {len(control_guidance_start)} elements but there are {len(self.controlnet.nets)} controlnets available. Make sure to provide {len(self.controlnet.nets)}."
|
731 |
+
)
|
732 |
+
|
733 |
+
for start, end in zip(control_guidance_start, control_guidance_end):
|
734 |
+
if start >= end:
|
735 |
+
raise ValueError(
|
736 |
+
f"control guidance start: {start} cannot be larger or equal to control guidance end: {end}."
|
737 |
+
)
|
738 |
+
if start < 0.0:
|
739 |
+
raise ValueError(f"control guidance start: {start} can't be smaller than 0.")
|
740 |
+
if end > 1.0:
|
741 |
+
raise ValueError(f"control guidance end: {end} can't be larger than 1.0.")
|
742 |
+
|
743 |
+
if ip_adapter_image is not None and ip_adapter_image_embeds is not None:
|
744 |
+
raise ValueError(
|
745 |
+
"Provide either `ip_adapter_image` or `ip_adapter_image_embeds`. Cannot leave both `ip_adapter_image` and `ip_adapter_image_embeds` defined."
|
746 |
+
)
|
747 |
+
|
748 |
+
if ip_adapter_image_embeds is not None:
|
749 |
+
if not isinstance(ip_adapter_image_embeds, list):
|
750 |
+
raise ValueError(
|
751 |
+
f"`ip_adapter_image_embeds` has to be of type `list` but is {type(ip_adapter_image_embeds)}"
|
752 |
+
)
|
753 |
+
elif ip_adapter_image_embeds[0].ndim not in [3, 4]:
|
754 |
+
raise ValueError(
|
755 |
+
f"`ip_adapter_image_embeds` has to be a list of 3D or 4D tensors but is {ip_adapter_image_embeds[0].ndim}D"
|
756 |
+
)
|
757 |
+
|
758 |
+
def check_image(self, image, prompt, prompt_embeds):
|
759 |
+
image_is_pil = isinstance(image, PIL.Image.Image)
|
760 |
+
image_is_tensor = isinstance(image, torch.Tensor)
|
761 |
+
image_is_np = isinstance(image, np.ndarray)
|
762 |
+
image_is_pil_list = isinstance(image, list) and isinstance(image[0], PIL.Image.Image)
|
763 |
+
image_is_tensor_list = isinstance(image, list) and isinstance(image[0], torch.Tensor)
|
764 |
+
image_is_np_list = isinstance(image, list) and isinstance(image[0], np.ndarray)
|
765 |
+
|
766 |
+
if (
|
767 |
+
not image_is_pil
|
768 |
+
and not image_is_tensor
|
769 |
+
and not image_is_np
|
770 |
+
and not image_is_pil_list
|
771 |
+
and not image_is_tensor_list
|
772 |
+
and not image_is_np_list
|
773 |
+
):
|
774 |
+
raise TypeError(
|
775 |
+
f"image must be passed and be one of PIL image, numpy array, torch tensor, list of PIL images, list of numpy arrays or list of torch tensors, but is {type(image)}"
|
776 |
+
)
|
777 |
+
|
778 |
+
if image_is_pil:
|
779 |
+
image_batch_size = 1
|
780 |
+
else:
|
781 |
+
image_batch_size = len(image)
|
782 |
+
|
783 |
+
if prompt is not None and isinstance(prompt, str):
|
784 |
+
prompt_batch_size = 1
|
785 |
+
elif prompt is not None and isinstance(prompt, list):
|
786 |
+
prompt_batch_size = len(prompt)
|
787 |
+
elif prompt_embeds is not None:
|
788 |
+
prompt_batch_size = prompt_embeds.shape[0]
|
789 |
+
|
790 |
+
if image_batch_size != 1 and image_batch_size != prompt_batch_size:
|
791 |
+
raise ValueError(
|
792 |
+
f"If image batch size is not 1, image batch size must be same as prompt batch size. image batch size: {image_batch_size}, prompt batch size: {prompt_batch_size}"
|
793 |
+
)
|
794 |
+
|
795 |
+
def prepare_image(
|
796 |
+
self,
|
797 |
+
image,
|
798 |
+
width,
|
799 |
+
height,
|
800 |
+
batch_size,
|
801 |
+
num_images_per_prompt,
|
802 |
+
device,
|
803 |
+
dtype,
|
804 |
+
do_classifier_free_guidance=False,
|
805 |
+
guess_mode=False,
|
806 |
+
):
|
807 |
+
image = self.control_image_processor.preprocess(image, height=height, width=width).to(dtype=torch.float32)
|
808 |
+
image_batch_size = image.shape[0]
|
809 |
+
|
810 |
+
if image_batch_size == 1:
|
811 |
+
repeat_by = batch_size
|
812 |
+
else:
|
813 |
+
# image batch size is the same as prompt batch size
|
814 |
+
repeat_by = num_images_per_prompt
|
815 |
+
|
816 |
+
image = image.repeat_interleave(repeat_by, dim=0)
|
817 |
+
|
818 |
+
image = image.to(device=device, dtype=dtype)
|
819 |
+
|
820 |
+
if do_classifier_free_guidance and not guess_mode:
|
821 |
+
image = torch.cat([image] * 2)
|
822 |
+
|
823 |
+
return image
|
824 |
+
|
825 |
+
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
|
826 |
+
def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
|
827 |
+
shape = (
|
828 |
+
batch_size,
|
829 |
+
num_channels_latents,
|
830 |
+
int(height) // self.vae_scale_factor,
|
831 |
+
int(width) // self.vae_scale_factor,
|
832 |
+
)
|
833 |
+
if isinstance(generator, list) and len(generator) != batch_size:
|
834 |
+
raise ValueError(
|
835 |
+
f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
|
836 |
+
f" size of {batch_size}. Make sure the batch size matches the length of the generators."
|
837 |
+
)
|
838 |
+
|
839 |
+
if latents is None:
|
840 |
+
latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
|
841 |
+
else:
|
842 |
+
latents = latents.to(device)
|
843 |
+
|
844 |
+
# scale the initial noise by the standard deviation required by the scheduler
|
845 |
+
latents = latents * self.scheduler.init_noise_sigma
|
846 |
+
return latents
|
847 |
+
|
848 |
+
# Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding
|
849 |
+
def get_guidance_scale_embedding(
|
850 |
+
self, w: torch.Tensor, embedding_dim: int = 512, dtype: torch.dtype = torch.float32
|
851 |
+
) -> torch.Tensor:
|
852 |
+
"""
|
853 |
+
See https://github.com/google-research/vdm/blob/dc27b98a554f65cdc654b800da5aa1846545d41b/model_vdm.py#L298
|
854 |
+
|
855 |
+
Args:
|
856 |
+
w (`torch.Tensor`):
|
857 |
+
Generate embedding vectors with a specified guidance scale to subsequently enrich timestep embeddings.
|
858 |
+
embedding_dim (`int`, *optional*, defaults to 512):
|
859 |
+
Dimension of the embeddings to generate.
|
860 |
+
dtype (`torch.dtype`, *optional*, defaults to `torch.float32`):
|
861 |
+
Data type of the generated embeddings.
|
862 |
+
|
863 |
+
Returns:
|
864 |
+
`torch.Tensor`: Embedding vectors with shape `(len(w), embedding_dim)`.
|
865 |
+
"""
|
866 |
+
assert len(w.shape) == 1
|
867 |
+
w = w * 1000.0
|
868 |
+
|
869 |
+
half_dim = embedding_dim // 2
|
870 |
+
emb = torch.log(torch.tensor(10000.0)) / (half_dim - 1)
|
871 |
+
emb = torch.exp(torch.arange(half_dim, dtype=dtype) * -emb)
|
872 |
+
emb = w.to(dtype)[:, None] * emb[None, :]
|
873 |
+
emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
|
874 |
+
if embedding_dim % 2 == 1: # zero pad
|
875 |
+
emb = torch.nn.functional.pad(emb, (0, 1))
|
876 |
+
assert emb.shape == (w.shape[0], embedding_dim)
|
877 |
+
return emb
|
878 |
+
|
879 |
+
@property
|
880 |
+
def guidance_scale(self):
|
881 |
+
return self._guidance_scale
|
882 |
+
|
883 |
+
@property
|
884 |
+
def clip_skip(self):
|
885 |
+
return self._clip_skip
|
886 |
+
|
887 |
+
# here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
|
888 |
+
# of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
|
889 |
+
# corresponds to doing no classifier free guidance.
|
890 |
+
@property
|
891 |
+
def do_classifier_free_guidance(self):
|
892 |
+
return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
|
893 |
+
|
894 |
+
@property
|
895 |
+
def cross_attention_kwargs(self):
|
896 |
+
return self._cross_attention_kwargs
|
897 |
+
|
898 |
+
@property
|
899 |
+
def num_timesteps(self):
|
900 |
+
return self._num_timesteps
|
901 |
+
|
902 |
+
@torch.no_grad()
|
903 |
+
@replace_example_docstring(EXAMPLE_DOC_STRING)
|
904 |
+
def __call__(
|
905 |
+
self,
|
906 |
+
prompt: Union[str, List[str]] = None,
|
907 |
+
image: PipelineImageInput = None,
|
908 |
+
height: Optional[int] = None,
|
909 |
+
width: Optional[int] = None,
|
910 |
+
num_inference_steps: int = 50,
|
911 |
+
timesteps: List[int] = None,
|
912 |
+
sigmas: List[float] = None,
|
913 |
+
guidance_scale: float = 7.5,
|
914 |
+
negative_prompt: Optional[Union[str, List[str]]] = None,
|
915 |
+
num_images_per_prompt: Optional[int] = 1,
|
916 |
+
eta: float = 0.0,
|
917 |
+
generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
|
918 |
+
latents: Optional[torch.Tensor] = None,
|
919 |
+
prompt_embeds: Optional[torch.Tensor] = None,
|
920 |
+
negative_prompt_embeds: Optional[torch.Tensor] = None,
|
921 |
+
ip_adapter_image: Optional[PipelineImageInput] = None,
|
922 |
+
ip_adapter_image_embeds: Optional[List[torch.Tensor]] = None,
|
923 |
+
output_type: Optional[str] = "pil",
|
924 |
+
return_dict: bool = True,
|
925 |
+
cross_attention_kwargs: Optional[Dict[str, Any]] = None,
|
926 |
+
controlnet_conditioning_scale: Union[float, List[float]] = 1.0,
|
927 |
+
guess_mode: bool = False,
|
928 |
+
control_guidance_start: Union[float, List[float]] = 0.0,
|
929 |
+
control_guidance_end: Union[float, List[float]] = 1.0,
|
930 |
+
clip_skip: Optional[int] = None,
|
931 |
+
callback_on_step_end: Optional[
|
932 |
+
Union[Callable[[int, int, Dict], None], PipelineCallback, MultiPipelineCallbacks]
|
933 |
+
] = None,
|
934 |
+
callback_on_step_end_tensor_inputs: List[str] = ["latents"],
|
935 |
+
**kwargs,
|
936 |
+
):
|
937 |
+
r"""
|
938 |
+
The call function to the pipeline for generation.
|
939 |
+
|
940 |
+
Args:
|
941 |
+
prompt (`str` or `List[str]`, *optional*):
|
942 |
+
The prompt or prompts to guide image generation. If not defined, you need to pass `prompt_embeds`.
|
943 |
+
image (`torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.Tensor]`, `List[PIL.Image.Image]`, `List[np.ndarray]`,:
|
944 |
+
`List[List[torch.Tensor]]`, `List[List[np.ndarray]]` or `List[List[PIL.Image.Image]]`):
|
945 |
+
The ControlNet input condition to provide guidance to the `unet` for generation. If the type is
|
946 |
+
specified as `torch.Tensor`, it is passed to ControlNet as is. `PIL.Image.Image` can also be accepted
|
947 |
+
as an image. The dimensions of the output image defaults to `image`'s dimensions. If height and/or
|
948 |
+
width are passed, `image` is resized accordingly. If multiple ControlNets are specified in `init`,
|
949 |
+
images must be passed as a list such that each element of the list can be correctly batched for input
|
950 |
+
to a single ControlNet. When `prompt` is a list, and if a list of images is passed for a single
|
951 |
+
ControlNet, each will be paired with each prompt in the `prompt` list. This also applies to multiple
|
952 |
+
ControlNets, where a list of image lists can be passed to batch for each prompt and each ControlNet.
|
953 |
+
height (`int`, *optional*, defaults to `self.unet.config.sample_size * self.vae_scale_factor`):
|
954 |
+
The height in pixels of the generated image.
|
955 |
+
width (`int`, *optional*, defaults to `self.unet.config.sample_size * self.vae_scale_factor`):
|
956 |
+
The width in pixels of the generated image.
|
957 |
+
num_inference_steps (`int`, *optional*, defaults to 50):
|
958 |
+
The number of denoising steps. More denoising steps usually lead to a higher quality image at the
|
959 |
+
expense of slower inference.
|
960 |
+
timesteps (`List[int]`, *optional*):
|
961 |
+
Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument
|
962 |
+
in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is
|
963 |
+
passed will be used. Must be in descending order.
|
964 |
+
sigmas (`List[float]`, *optional*):
|
965 |
+
Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in
|
966 |
+
their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed
|
967 |
+
will be used.
|
968 |
+
guidance_scale (`float`, *optional*, defaults to 7.5):
|
969 |
+
A higher guidance scale value encourages the model to generate images closely linked to the text
|
970 |
+
`prompt` at the expense of lower image quality. Guidance scale is enabled when `guidance_scale > 1`.
|
971 |
+
negative_prompt (`str` or `List[str]`, *optional*):
|
972 |
+
The prompt or prompts to guide what to not include in image generation. If not defined, you need to
|
973 |
+
pass `negative_prompt_embeds` instead. Ignored when not using guidance (`guidance_scale < 1`).
|
974 |
+
num_images_per_prompt (`int`, *optional*, defaults to 1):
|
975 |
+
The number of images to generate per prompt.
|
976 |
+
eta (`float`, *optional*, defaults to 0.0):
|
977 |
+
Corresponds to parameter eta (η) from the [DDIM](https://arxiv.org/abs/2010.02502) paper. Only applies
|
978 |
+
to the [`~schedulers.DDIMScheduler`], and is ignored in other schedulers.
|
979 |
+
generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
|
980 |
+
A [`torch.Generator`](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make
|
981 |
+
generation deterministic.
|
982 |
+
latents (`torch.Tensor`, *optional*):
|
983 |
+
Pre-generated noisy latents sampled from a Gaussian distribution, to be used as inputs for image
|
984 |
+
generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
|
985 |
+
tensor is generated by sampling using the supplied random `generator`.
|
986 |
+
prompt_embeds (`torch.Tensor`, *optional*):
|
987 |
+
Pre-generated text embeddings. Can be used to easily tweak text inputs (prompt weighting). If not
|
988 |
+
provided, text embeddings are generated from the `prompt` input argument.
|
989 |
+
negative_prompt_embeds (`torch.Tensor`, *optional*):
|
990 |
+
Pre-generated negative text embeddings. Can be used to easily tweak text inputs (prompt weighting). If
|
991 |
+
not provided, `negative_prompt_embeds` are generated from the `negative_prompt` input argument.
|
992 |
+
ip_adapter_image: (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
|
993 |
+
ip_adapter_image_embeds (`List[torch.Tensor]`, *optional*):
|
994 |
+
Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of
|
995 |
+
IP-adapters. Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. It should
|
996 |
+
contain the negative image embedding if `do_classifier_free_guidance` is set to `True`. If not
|
997 |
+
provided, embeddings are computed from the `ip_adapter_image` input argument.
|
998 |
+
output_type (`str`, *optional*, defaults to `"pil"`):
|
999 |
+
The output format of the generated image. Choose between `PIL.Image` or `np.array`.
|
1000 |
+
return_dict (`bool`, *optional*, defaults to `True`):
|
1001 |
+
Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
|
1002 |
+
plain tuple.
|
1003 |
+
callback (`Callable`, *optional*):
|
1004 |
+
A function that calls every `callback_steps` steps during inference. The function is called with the
|
1005 |
+
following arguments: `callback(step: int, timestep: int, latents: torch.Tensor)`.
|
1006 |
+
callback_steps (`int`, *optional*, defaults to 1):
|
1007 |
+
The frequency at which the `callback` function is called. If not specified, the callback is called at
|
1008 |
+
every step.
|
1009 |
+
cross_attention_kwargs (`dict`, *optional*):
|
1010 |
+
A kwargs dictionary that if specified is passed along to the [`AttentionProcessor`] as defined in
|
1011 |
+
[`self.processor`](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
|
1012 |
+
controlnet_conditioning_scale (`float` or `List[float]`, *optional*, defaults to 1.0):
|
1013 |
+
The outputs of the ControlNet are multiplied by `controlnet_conditioning_scale` before they are added
|
1014 |
+
to the residual in the original `unet`. If multiple ControlNets are specified in `init`, you can set
|
1015 |
+
the corresponding scale as a list.
|
1016 |
+
guess_mode (`bool`, *optional*, defaults to `False`):
|
1017 |
+
The ControlNet encoder tries to recognize the content of the input image even if you remove all
|
1018 |
+
prompts. A `guidance_scale` value between 3.0 and 5.0 is recommended.
|
1019 |
+
control_guidance_start (`float` or `List[float]`, *optional*, defaults to 0.0):
|
1020 |
+
The percentage of total steps at which the ControlNet starts applying.
|
1021 |
+
control_guidance_end (`float` or `List[float]`, *optional*, defaults to 1.0):
|
1022 |
+
The percentage of total steps at which the ControlNet stops applying.
|
1023 |
+
clip_skip (`int`, *optional*):
|
1024 |
+
Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
|
1025 |
+
the output of the pre-final layer will be used for computing the prompt embeddings.
|
1026 |
+
callback_on_step_end (`Callable`, `PipelineCallback`, `MultiPipelineCallbacks`, *optional*):
|
1027 |
+
A function or a subclass of `PipelineCallback` or `MultiPipelineCallbacks` that is called at the end of
|
1028 |
+
each denoising step during the inference. with the following arguments: `callback_on_step_end(self:
|
1029 |
+
DiffusionPipeline, step: int, timestep: int, callback_kwargs: Dict)`. `callback_kwargs` will include a
|
1030 |
+
list of all tensors as specified by `callback_on_step_end_tensor_inputs`.
|
1031 |
+
callback_on_step_end_tensor_inputs (`List`, *optional*):
|
1032 |
+
The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list
|
1033 |
+
will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the
|
1034 |
+
`._callback_tensor_inputs` attribute of your pipeline class.
|
1035 |
+
|
1036 |
+
Examples:
|
1037 |
+
|
1038 |
+
Returns:
|
1039 |
+
[`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`:
|
1040 |
+
If `return_dict` is `True`, [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] is returned,
|
1041 |
+
otherwise a `tuple` is returned where the first element is a list with the generated images and the
|
1042 |
+
second element is a list of `bool`s indicating whether the corresponding generated image contains
|
1043 |
+
"not-safe-for-work" (nsfw) content.
|
1044 |
+
"""
|
1045 |
+
|
1046 |
+
callback = kwargs.pop("callback", None)
|
1047 |
+
callback_steps = kwargs.pop("callback_steps", None)
|
1048 |
+
|
1049 |
+
if callback is not None:
|
1050 |
+
deprecate(
|
1051 |
+
"callback",
|
1052 |
+
"1.0.0",
|
1053 |
+
"Passing `callback` as an input argument to `__call__` is deprecated, consider using `callback_on_step_end`",
|
1054 |
+
)
|
1055 |
+
if callback_steps is not None:
|
1056 |
+
deprecate(
|
1057 |
+
"callback_steps",
|
1058 |
+
"1.0.0",
|
1059 |
+
"Passing `callback_steps` as an input argument to `__call__` is deprecated, consider using `callback_on_step_end`",
|
1060 |
+
)
|
1061 |
+
|
1062 |
+
if isinstance(callback_on_step_end, (PipelineCallback, MultiPipelineCallbacks)):
|
1063 |
+
callback_on_step_end_tensor_inputs = callback_on_step_end.tensor_inputs
|
1064 |
+
|
1065 |
+
controlnet = self.controlnet._orig_mod if is_compiled_module(self.controlnet) else self.controlnet
|
1066 |
+
|
1067 |
+
# align format for control guidance
|
1068 |
+
if not isinstance(control_guidance_start, list) and isinstance(control_guidance_end, list):
|
1069 |
+
control_guidance_start = len(control_guidance_end) * [control_guidance_start]
|
1070 |
+
elif not isinstance(control_guidance_end, list) and isinstance(control_guidance_start, list):
|
1071 |
+
control_guidance_end = len(control_guidance_start) * [control_guidance_end]
|
1072 |
+
elif not isinstance(control_guidance_start, list) and not isinstance(control_guidance_end, list):
|
1073 |
+
mult = len(controlnet.nets) if isinstance(controlnet, MultiControlNetModel) else 1
|
1074 |
+
control_guidance_start, control_guidance_end = (
|
1075 |
+
mult * [control_guidance_start],
|
1076 |
+
mult * [control_guidance_end],
|
1077 |
+
)
|
1078 |
+
|
1079 |
+
# 1. Check inputs. Raise error if not correct
|
1080 |
+
self.check_inputs(
|
1081 |
+
prompt,
|
1082 |
+
image,
|
1083 |
+
callback_steps,
|
1084 |
+
negative_prompt,
|
1085 |
+
prompt_embeds,
|
1086 |
+
negative_prompt_embeds,
|
1087 |
+
ip_adapter_image,
|
1088 |
+
ip_adapter_image_embeds,
|
1089 |
+
controlnet_conditioning_scale,
|
1090 |
+
control_guidance_start,
|
1091 |
+
control_guidance_end,
|
1092 |
+
callback_on_step_end_tensor_inputs,
|
1093 |
+
)
|
1094 |
+
|
1095 |
+
self._guidance_scale = guidance_scale
|
1096 |
+
self._clip_skip = clip_skip
|
1097 |
+
self._cross_attention_kwargs = cross_attention_kwargs
|
1098 |
+
|
1099 |
+
# 2. Define call parameters
|
1100 |
+
if prompt is not None and isinstance(prompt, str):
|
1101 |
+
batch_size = 1
|
1102 |
+
elif prompt is not None and isinstance(prompt, list):
|
1103 |
+
batch_size = len(prompt)
|
1104 |
+
else:
|
1105 |
+
batch_size = prompt_embeds.shape[0]
|
1106 |
+
|
1107 |
+
device = self._execution_device
|
1108 |
+
|
1109 |
+
if isinstance(controlnet, MultiControlNetModel) and isinstance(controlnet_conditioning_scale, float):
|
1110 |
+
controlnet_conditioning_scale = [controlnet_conditioning_scale] * len(controlnet.nets)
|
1111 |
+
|
1112 |
+
global_pool_conditions = (
|
1113 |
+
controlnet.config.global_pool_conditions
|
1114 |
+
if isinstance(controlnet, ControlNetModel)
|
1115 |
+
else controlnet.nets[0].config.global_pool_conditions
|
1116 |
+
)
|
1117 |
+
guess_mode = guess_mode or global_pool_conditions
|
1118 |
+
|
1119 |
+
# 3. Encode input prompt
|
1120 |
+
text_encoder_lora_scale = (
|
1121 |
+
self.cross_attention_kwargs.get("scale", None) if self.cross_attention_kwargs is not None else None
|
1122 |
+
)
|
1123 |
+
prompt_embeds, negative_prompt_embeds = self.encode_prompt(
|
1124 |
+
prompt,
|
1125 |
+
device,
|
1126 |
+
num_images_per_prompt,
|
1127 |
+
self.do_classifier_free_guidance,
|
1128 |
+
negative_prompt,
|
1129 |
+
prompt_embeds=prompt_embeds,
|
1130 |
+
negative_prompt_embeds=negative_prompt_embeds,
|
1131 |
+
lora_scale=text_encoder_lora_scale,
|
1132 |
+
clip_skip=self.clip_skip,
|
1133 |
+
)
|
1134 |
+
# For classifier free guidance, we need to do two forward passes.
|
1135 |
+
# Here we concatenate the unconditional and text embeddings into a single batch
|
1136 |
+
# to avoid doing two forward passes
|
1137 |
+
if self.do_classifier_free_guidance:
|
1138 |
+
prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds])
|
1139 |
+
|
1140 |
+
if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
|
1141 |
+
image_embeds = self.prepare_ip_adapter_image_embeds(
|
1142 |
+
ip_adapter_image,
|
1143 |
+
ip_adapter_image_embeds,
|
1144 |
+
device,
|
1145 |
+
batch_size * num_images_per_prompt,
|
1146 |
+
self.do_classifier_free_guidance,
|
1147 |
+
)
|
1148 |
+
|
1149 |
+
# 4. Prepare image
|
1150 |
+
if isinstance(controlnet, ControlNetModel):
|
1151 |
+
image = self.prepare_image(
|
1152 |
+
image=image,
|
1153 |
+
width=width,
|
1154 |
+
height=height,
|
1155 |
+
batch_size=batch_size * num_images_per_prompt,
|
1156 |
+
num_images_per_prompt=num_images_per_prompt,
|
1157 |
+
device=device,
|
1158 |
+
dtype=controlnet.dtype,
|
1159 |
+
do_classifier_free_guidance=self.do_classifier_free_guidance,
|
1160 |
+
guess_mode=guess_mode,
|
1161 |
+
)
|
1162 |
+
height, width = image.shape[-2:]
|
1163 |
+
elif isinstance(controlnet, MultiControlNetModel):
|
1164 |
+
images = []
|
1165 |
+
|
1166 |
+
# Nested lists as ControlNet condition
|
1167 |
+
if isinstance(image[0], list):
|
1168 |
+
# Transpose the nested image list
|
1169 |
+
image = [list(t) for t in zip(*image)]
|
1170 |
+
|
1171 |
+
for image_ in image:
|
1172 |
+
image_ = self.prepare_image(
|
1173 |
+
image=image_,
|
1174 |
+
width=width,
|
1175 |
+
height=height,
|
1176 |
+
batch_size=batch_size * num_images_per_prompt,
|
1177 |
+
num_images_per_prompt=num_images_per_prompt,
|
1178 |
+
device=device,
|
1179 |
+
dtype=controlnet.dtype,
|
1180 |
+
do_classifier_free_guidance=self.do_classifier_free_guidance,
|
1181 |
+
guess_mode=guess_mode,
|
1182 |
+
)
|
1183 |
+
|
1184 |
+
images.append(image_)
|
1185 |
+
|
1186 |
+
image = images
|
1187 |
+
height, width = image[0].shape[-2:]
|
1188 |
+
else:
|
1189 |
+
assert False
|
1190 |
+
|
1191 |
+
# 5. Prepare timesteps
|
1192 |
+
timesteps, num_inference_steps = retrieve_timesteps(
|
1193 |
+
self.scheduler, num_inference_steps, device, timesteps, sigmas
|
1194 |
+
)
|
1195 |
+
self._num_timesteps = len(timesteps)
|
1196 |
+
|
1197 |
+
# 6. Prepare latent variables
|
1198 |
+
num_channels_latents = self.unet.config.in_channels
|
1199 |
+
latents = self.prepare_latents(
|
1200 |
+
batch_size * num_images_per_prompt,
|
1201 |
+
num_channels_latents,
|
1202 |
+
height,
|
1203 |
+
width,
|
1204 |
+
prompt_embeds.dtype,
|
1205 |
+
device,
|
1206 |
+
generator,
|
1207 |
+
latents,
|
1208 |
+
)
|
1209 |
+
|
1210 |
+
# 6.5 Optionally get Guidance Scale Embedding
|
1211 |
+
timestep_cond = None
|
1212 |
+
if self.unet.config.time_cond_proj_dim is not None:
|
1213 |
+
guidance_scale_tensor = torch.tensor(self.guidance_scale - 1).repeat(batch_size * num_images_per_prompt)
|
1214 |
+
timestep_cond = self.get_guidance_scale_embedding(
|
1215 |
+
guidance_scale_tensor, embedding_dim=self.unet.config.time_cond_proj_dim
|
1216 |
+
).to(device=device, dtype=latents.dtype)
|
1217 |
+
|
1218 |
+
# 7. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline
|
1219 |
+
extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)
|
1220 |
+
|
1221 |
+
# 7.1 Add image embeds for IP-Adapter
|
1222 |
+
added_cond_kwargs = (
|
1223 |
+
{"image_embeds": image_embeds}
|
1224 |
+
if ip_adapter_image is not None or ip_adapter_image_embeds is not None
|
1225 |
+
else None
|
1226 |
+
)
|
1227 |
+
|
1228 |
+
# 7.2 Create tensor stating which controlnets to keep
|
1229 |
+
controlnet_keep = []
|
1230 |
+
for i in range(len(timesteps)):
|
1231 |
+
keeps = [
|
1232 |
+
1.0 - float(i / len(timesteps) < s or (i + 1) / len(timesteps) > e)
|
1233 |
+
for s, e in zip(control_guidance_start, control_guidance_end)
|
1234 |
+
]
|
1235 |
+
controlnet_keep.append(keeps[0] if isinstance(controlnet, ControlNetModel) else keeps)
|
1236 |
+
|
1237 |
+
# 8. Denoising loop
|
1238 |
+
num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
|
1239 |
+
is_unet_compiled = is_compiled_module(self.unet)
|
1240 |
+
is_controlnet_compiled = is_compiled_module(self.controlnet)
|
1241 |
+
is_torch_higher_equal_2_1 = is_torch_version(">=", "2.1")
|
1242 |
+
with self.progress_bar(total=num_inference_steps) as progress_bar:
|
1243 |
+
for i, t in enumerate(timesteps):
|
1244 |
+
# Relevant thread:
|
1245 |
+
# https://dev-discuss.pytorch.org/t/cudagraphs-in-pytorch-2-0/1428
|
1246 |
+
if (is_unet_compiled and is_controlnet_compiled) and is_torch_higher_equal_2_1:
|
1247 |
+
torch._inductor.cudagraph_mark_step_begin()
|
1248 |
+
# expand the latents if we are doing classifier free guidance
|
1249 |
+
latent_model_input = torch.cat([latents] * 2) if self.do_classifier_free_guidance else latents
|
1250 |
+
latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
|
1251 |
+
|
1252 |
+
# controlnet(s) inference
|
1253 |
+
if guess_mode and self.do_classifier_free_guidance:
|
1254 |
+
# Infer ControlNet only for the conditional batch.
|
1255 |
+
control_model_input = latents
|
1256 |
+
control_model_input = self.scheduler.scale_model_input(control_model_input, t)
|
1257 |
+
controlnet_prompt_embeds = prompt_embeds.chunk(2)[1]
|
1258 |
+
else:
|
1259 |
+
control_model_input = latent_model_input
|
1260 |
+
controlnet_prompt_embeds = prompt_embeds
|
1261 |
+
|
1262 |
+
if isinstance(controlnet_keep[i], list):
|
1263 |
+
cond_scale = [c * s for c, s in zip(controlnet_conditioning_scale, controlnet_keep[i])]
|
1264 |
+
else:
|
1265 |
+
controlnet_cond_scale = controlnet_conditioning_scale
|
1266 |
+
if isinstance(controlnet_cond_scale, list):
|
1267 |
+
controlnet_cond_scale = controlnet_cond_scale[0]
|
1268 |
+
cond_scale = controlnet_cond_scale * controlnet_keep[i]
|
1269 |
+
|
1270 |
+
down_block_res_samples, mid_block_res_sample = self.controlnet(
|
1271 |
+
control_model_input,
|
1272 |
+
t,
|
1273 |
+
encoder_hidden_states=controlnet_prompt_embeds,
|
1274 |
+
controlnet_cond=image,
|
1275 |
+
conditioning_scale=cond_scale,
|
1276 |
+
guess_mode=guess_mode,
|
1277 |
+
return_dict=False,
|
1278 |
+
)
|
1279 |
+
|
1280 |
+
if guess_mode and self.do_classifier_free_guidance:
|
1281 |
+
# Infered ControlNet only for the conditional batch.
|
1282 |
+
# To apply the output of ControlNet to both the unconditional and conditional batches,
|
1283 |
+
# add 0 to the unconditional batch to keep it unchanged.
|
1284 |
+
down_block_res_samples = [torch.cat([torch.zeros_like(d), d]) for d in down_block_res_samples]
|
1285 |
+
mid_block_res_sample = torch.cat([torch.zeros_like(mid_block_res_sample), mid_block_res_sample])
|
1286 |
+
|
1287 |
+
# predict the noise residual
|
1288 |
+
noise_pred = self.unet(
|
1289 |
+
latent_model_input,
|
1290 |
+
t,
|
1291 |
+
encoder_hidden_states=prompt_embeds,
|
1292 |
+
timestep_cond=timestep_cond,
|
1293 |
+
cross_attention_kwargs=self.cross_attention_kwargs,
|
1294 |
+
down_block_additional_residuals=down_block_res_samples,
|
1295 |
+
mid_block_additional_residual=mid_block_res_sample,
|
1296 |
+
added_cond_kwargs=added_cond_kwargs,
|
1297 |
+
return_dict=False,
|
1298 |
+
)[0]
|
1299 |
+
|
1300 |
+
# perform guidance
|
1301 |
+
if self.do_classifier_free_guidance:
|
1302 |
+
noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
|
1303 |
+
noise_pred = noise_pred_uncond + self.guidance_scale * (noise_pred_text - noise_pred_uncond)
|
1304 |
+
|
1305 |
+
# compute the previous noisy sample x_t -> x_t-1
|
1306 |
+
latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
|
1307 |
+
|
1308 |
+
if callback_on_step_end is not None:
|
1309 |
+
callback_kwargs = {}
|
1310 |
+
for k in callback_on_step_end_tensor_inputs:
|
1311 |
+
callback_kwargs[k] = locals()[k]
|
1312 |
+
callback_outputs = callback_on_step_end(self, i, t, callback_kwargs)
|
1313 |
+
|
1314 |
+
latents = callback_outputs.pop("latents", latents)
|
1315 |
+
prompt_embeds = callback_outputs.pop("prompt_embeds", prompt_embeds)
|
1316 |
+
negative_prompt_embeds = callback_outputs.pop("negative_prompt_embeds", negative_prompt_embeds)
|
1317 |
+
|
1318 |
+
# call the callback, if provided
|
1319 |
+
if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
|
1320 |
+
progress_bar.update()
|
1321 |
+
if callback is not None and i % callback_steps == 0:
|
1322 |
+
step_idx = i // getattr(self.scheduler, "order", 1)
|
1323 |
+
callback(step_idx, t, latents)
|
1324 |
+
|
1325 |
+
# If we do sequential model offloading, let's offload unet and controlnet
|
1326 |
+
# manually for max memory savings
|
1327 |
+
if hasattr(self, "final_offload_hook") and self.final_offload_hook is not None:
|
1328 |
+
self.unet.to("cpu")
|
1329 |
+
self.controlnet.to("cpu")
|
1330 |
+
torch.cuda.empty_cache()
|
1331 |
+
|
1332 |
+
if not output_type == "latent":
|
1333 |
+
image = self.vae.decode(latents / self.vae.config.scaling_factor, return_dict=False, generator=generator)[
|
1334 |
+
0
|
1335 |
+
]
|
1336 |
+
image, has_nsfw_concept = self.run_safety_checker(image, device, prompt_embeds.dtype)
|
1337 |
+
else:
|
1338 |
+
image = latents
|
1339 |
+
has_nsfw_concept = None
|
1340 |
+
|
1341 |
+
if has_nsfw_concept is None:
|
1342 |
+
do_denormalize = [True] * image.shape[0]
|
1343 |
+
else:
|
1344 |
+
do_denormalize = [not has_nsfw for has_nsfw in has_nsfw_concept]
|
1345 |
+
|
1346 |
+
image = self.image_processor.postprocess(image, output_type=output_type, do_denormalize=do_denormalize)
|
1347 |
+
|
1348 |
+
# Offload all models
|
1349 |
+
self.maybe_free_model_hooks()
|
1350 |
+
|
1351 |
+
if not return_dict:
|
1352 |
+
return (image, has_nsfw_concept)
|
1353 |
+
|
1354 |
+
return StableDiffusionPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept)
|