Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -1,70 +1,119 @@
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
import numpy as np
|
3 |
-
import
|
4 |
-
from diffusers import DiffusionPipeline
|
5 |
-
from diffusers import StableDiffusionXLPipeline, DPMSolverSinglestepScheduler
|
6 |
-
import torch
|
7 |
import spaces
|
|
|
|
|
8 |
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
pipe = StableDiffusionXLPipeline.from_pretrained("sd-community/sdxl-flash")
|
13 |
-
pipe = pipe.to(device)
|
14 |
-
pipe.scheduler = DPMSolverSinglestepScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
|
15 |
|
16 |
MAX_SEED = np.iinfo(np.int32).max
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
if randomize_seed:
|
31 |
-
seed = random.randint(0, MAX_SEED)
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
43 |
|
44 |
examples = [
|
45 |
-
"
|
46 |
-
"
|
47 |
-
"
|
48 |
-
"
|
49 |
-
"
|
|
|
50 |
]
|
51 |
|
52 |
css = '''
|
53 |
.gradio-container{max-width: 560px !important}
|
54 |
h1{text-align:center}
|
55 |
-
footer {
|
56 |
-
visibility: hidden
|
57 |
-
}
|
58 |
'''
|
59 |
-
|
60 |
-
|
61 |
-
with gr.
|
62 |
-
gr.Markdown("""# SDXL Flash
|
63 |
-
### Super fast text to Image Generator.
|
64 |
-
### <span style='color: red;'>You may change the steps from 5 to 8 or 10, if you didn't get satisfied results.
|
65 |
-
### First Image processing takes time then images generate faster.""")
|
66 |
with gr.Row():
|
67 |
-
|
68 |
prompt = gr.Text(
|
69 |
label="Prompt",
|
70 |
show_label=False,
|
@@ -72,79 +121,93 @@ with gr.Blocks(title="SDXL Flash", css=css) as demo:
|
|
72 |
placeholder="Enter your prompt",
|
73 |
container=False,
|
74 |
)
|
75 |
-
|
76 |
run_button = gr.Button("Run", scale=0)
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
negative_prompt = gr.Text(
|
83 |
label="Negative prompt",
|
84 |
-
max_lines=
|
85 |
-
lines=4,
|
86 |
placeholder="Enter a negative prompt",
|
87 |
-
value
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
88 |
)
|
89 |
-
|
90 |
-
|
91 |
-
label="
|
92 |
-
minimum=0,
|
93 |
-
maximum=
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
94 |
step=1,
|
95 |
-
value=
|
96 |
)
|
97 |
-
|
98 |
-
randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
|
99 |
-
|
100 |
-
with gr.Row():
|
101 |
-
|
102 |
-
width = gr.Slider(
|
103 |
-
label="Width",
|
104 |
-
minimum=256,
|
105 |
-
maximum=MAX_IMAGE_SIZE,
|
106 |
-
step=8,
|
107 |
-
value=1024,
|
108 |
-
)
|
109 |
-
|
110 |
-
height = gr.Slider(
|
111 |
-
label="Height",
|
112 |
-
minimum=256,
|
113 |
-
maximum=MAX_IMAGE_SIZE,
|
114 |
-
step=8,
|
115 |
-
value=1024,
|
116 |
-
)
|
117 |
-
|
118 |
-
with gr.Row():
|
119 |
-
|
120 |
-
guidance_scale = gr.Slider(
|
121 |
-
label="Guidance scale",
|
122 |
-
minimum=1.0,
|
123 |
-
maximum=6.0,
|
124 |
-
step=0.1,
|
125 |
-
value=3.0,
|
126 |
-
)
|
127 |
-
|
128 |
-
num_inference_steps = gr.Slider(
|
129 |
-
label="Number of inference steps",
|
130 |
-
minimum=1,
|
131 |
-
maximum=15,
|
132 |
-
step=1,
|
133 |
-
value=5,
|
134 |
-
)
|
135 |
-
|
136 |
-
gr.Examples(
|
137 |
-
examples = examples,
|
138 |
-
inputs = prompt,
|
139 |
-
outputs = result,
|
140 |
-
fn=infer,
|
141 |
-
cache_examples=True
|
142 |
-
)
|
143 |
|
144 |
-
|
145 |
-
|
146 |
-
inputs
|
147 |
-
outputs
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
148 |
)
|
149 |
|
150 |
-
|
|
|
|
1 |
+
import os
|
2 |
+
import random
|
3 |
+
import uuid
|
4 |
+
import json
|
5 |
+
|
6 |
import gradio as gr
|
7 |
import numpy as np
|
8 |
+
from PIL import Image
|
|
|
|
|
|
|
9 |
import spaces
|
10 |
+
import torch
|
11 |
+
from diffusers import DiffusionPipeline
|
12 |
|
13 |
+
DESCRIPTION = """# SDXL Flash"""
|
14 |
+
if not torch.cuda.is_available():
|
15 |
+
DESCRIPTION += "\n<p>Running on CPU 🥶 This demo may not work on CPU.</p>"
|
|
|
|
|
|
|
16 |
|
17 |
MAX_SEED = np.iinfo(np.int32).max
|
18 |
+
CACHE_EXAMPLES = torch.cuda.is_available() and os.getenv("CACHE_EXAMPLES", "0") == "1"
|
19 |
+
MAX_IMAGE_SIZE = int(os.getenv("MAX_IMAGE_SIZE", "4096"))
|
20 |
+
USE_TORCH_COMPILE = os.getenv("USE_TORCH_COMPILE", "0") == "1"
|
21 |
+
ENABLE_CPU_OFFLOAD = os.getenv("ENABLE_CPU_OFFLOAD", "0") == "1"
|
22 |
+
|
23 |
+
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
24 |
+
|
25 |
+
NUM_IMAGES_PER_PROMPT = 1
|
26 |
+
|
27 |
+
if torch.cuda.is_available():
|
28 |
+
pipe = DiffusionPipeline.from_pretrained(
|
29 |
+
"sd-community/sdxl-flash",
|
30 |
+
torch_dtype=torch.float16,
|
31 |
+
use_safetensors=True,
|
32 |
+
add_watermarker=False,
|
33 |
+
variant="fp16"
|
34 |
+
)
|
35 |
+
if ENABLE_CPU_OFFLOAD:
|
36 |
+
pipe.enable_model_cpu_offload()
|
37 |
+
else:
|
38 |
+
pipe.to(device)
|
39 |
+
print("Loaded on Device!")
|
40 |
+
|
41 |
+
if USE_TORCH_COMPILE:
|
42 |
+
pipe.unet = torch.compile(pipe.unet, mode="reduce-overhead", fullgraph=True)
|
43 |
+
pipe2.unet = torch.compile(pipe2.unet, mode="reduce-overhead", fullgraph=True)
|
44 |
+
print("Model Compiled!")
|
45 |
+
|
46 |
+
|
47 |
+
def save_image(img):
|
48 |
+
unique_name = str(uuid.uuid4()) + ".png"
|
49 |
+
img.save(unique_name)
|
50 |
+
return unique_name
|
51 |
+
|
52 |
+
|
53 |
+
def randomize_seed_fn(seed: int, randomize_seed: bool) -> int:
|
54 |
if randomize_seed:
|
55 |
+
seed = random.randint(0, MAX_SEED)
|
56 |
+
return seed
|
57 |
+
|
58 |
+
@spaces.GPU(enable_queue=False)
|
59 |
+
def generate(
|
60 |
+
prompt: str,
|
61 |
+
negative_prompt: str = "",
|
62 |
+
use_negative_prompt: bool = False,
|
63 |
+
seed: int = 0,
|
64 |
+
width: int = 1024,
|
65 |
+
height: int = 1024,
|
66 |
+
guidance_scale: float = 3,
|
67 |
+
num_inference_steps: int = 9,
|
68 |
+
randomize_seed: bool = False,
|
69 |
+
use_resolution_binning: bool = True,
|
70 |
+
progress=gr.Progress(track_tqdm=True),
|
71 |
+
):
|
72 |
+
pipe.to(device)
|
73 |
+
seed = int(randomize_seed_fn(seed, randomize_seed))
|
74 |
+
generator = torch.Generator().manual_seed(seed)
|
75 |
+
|
76 |
+
if not use_negative_prompt:
|
77 |
+
negative_prompt = "" # type: ignore
|
78 |
+
negative_prompt += default_negative
|
79 |
+
|
80 |
+
options = {
|
81 |
+
"prompt":prompt,
|
82 |
+
"negative_prompt":negative_prompt,
|
83 |
+
"width":width,
|
84 |
+
"height":height,
|
85 |
+
"guidance_scale":guidance_scale,
|
86 |
+
"num_inference_steps":num_inference_steps,
|
87 |
+
"generator":generator,
|
88 |
+
"num_images_per_prompt":NUM_IMAGES_PER_PROMPT,
|
89 |
+
"use_resolution_binning":use_resolution_binning,
|
90 |
+
"output_type":"pil",
|
91 |
+
|
92 |
+
}
|
93 |
+
|
94 |
+
images = pipe(**options).images
|
95 |
+
|
96 |
+
image_paths = [save_image(img) for img in images]
|
97 |
+
return image_paths, seed
|
98 |
+
|
99 |
|
100 |
examples = [
|
101 |
+
"neon holography crystal cat",
|
102 |
+
"a cat eating a piece of cheese",
|
103 |
+
"an astronaut riding a horse in space",
|
104 |
+
"a cartoon of a boy playing with a tiger",
|
105 |
+
"a cute robot artist painting on an easel, concept art",
|
106 |
+
#"a close up of a woman wearing a transparent, prismatic, elaborate nemeses headdress, over the should pose, brown skin-tone"
|
107 |
]
|
108 |
|
109 |
css = '''
|
110 |
.gradio-container{max-width: 560px !important}
|
111 |
h1{text-align:center}
|
|
|
|
|
|
|
112 |
'''
|
113 |
+
with gr.Blocks(css=css) as demo:
|
114 |
+
gr.Markdown(DESCRIPTION)
|
115 |
+
with gr.Group():
|
|
|
|
|
|
|
|
|
116 |
with gr.Row():
|
|
|
117 |
prompt = gr.Text(
|
118 |
label="Prompt",
|
119 |
show_label=False,
|
|
|
121 |
placeholder="Enter your prompt",
|
122 |
container=False,
|
123 |
)
|
|
|
124 |
run_button = gr.Button("Run", scale=0)
|
125 |
+
result = gr.Gallery(label="Result", columns=NUM_IMAGES_PER_PROMPT, show_label=False)
|
126 |
+
with gr.Accordion("Advanced options", open=False):
|
127 |
+
with gr.Row():
|
128 |
+
use_negative_prompt = gr.Checkbox(label="Use negative prompt", value=True)
|
|
|
129 |
negative_prompt = gr.Text(
|
130 |
label="Negative prompt",
|
131 |
+
max_lines=1,
|
|
|
132 |
placeholder="Enter a negative prompt",
|
133 |
+
value="(deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers:1.4), disconnected limbs, mutation, mutated, ugly, disgusting, blurry, amputation, NSFW",
|
134 |
+
visible=True,
|
135 |
+
)
|
136 |
+
seed = gr.Slider(
|
137 |
+
label="Seed",
|
138 |
+
minimum=0,
|
139 |
+
maximum=MAX_SEED,
|
140 |
+
step=1,
|
141 |
+
value=0,
|
142 |
+
)
|
143 |
+
randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
|
144 |
+
with gr.Row(visible=True):
|
145 |
+
width = gr.Slider(
|
146 |
+
label="Width",
|
147 |
+
minimum=256,
|
148 |
+
maximum=MAX_IMAGE_SIZE,
|
149 |
+
step=32,
|
150 |
+
value=1024,
|
151 |
+
)
|
152 |
+
height = gr.Slider(
|
153 |
+
label="Height",
|
154 |
+
minimum=256,
|
155 |
+
maximum=MAX_IMAGE_SIZE,
|
156 |
+
step=32,
|
157 |
+
value=1024,
|
158 |
)
|
159 |
+
with gr.Row():
|
160 |
+
guidance_scale = gr.Slider(
|
161 |
+
label="Guidance Scale",
|
162 |
+
minimum=0.1,
|
163 |
+
maximum=6,
|
164 |
+
step=0.1,
|
165 |
+
value=3.0,
|
166 |
+
)
|
167 |
+
num_inference_steps = gr.Slider(
|
168 |
+
label="Number of inference steps",
|
169 |
+
minimum=1,
|
170 |
+
maximum=15,
|
171 |
step=1,
|
172 |
+
value=5,
|
173 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
174 |
|
175 |
+
gr.Examples(
|
176 |
+
examples=examples,
|
177 |
+
inputs=prompt,
|
178 |
+
outputs=[result, seed],
|
179 |
+
fn=generate,
|
180 |
+
cache_examples=CACHE_EXAMPLES,
|
181 |
+
)
|
182 |
+
|
183 |
+
use_negative_prompt.change(
|
184 |
+
fn=lambda x: gr.update(visible=x),
|
185 |
+
inputs=use_negative_prompt,
|
186 |
+
outputs=negative_prompt,
|
187 |
+
api_name=False,
|
188 |
+
)
|
189 |
+
|
190 |
+
gr.on(
|
191 |
+
triggers=[
|
192 |
+
prompt.submit,
|
193 |
+
negative_prompt.submit,
|
194 |
+
run_button.click,
|
195 |
+
],
|
196 |
+
fn=generate,
|
197 |
+
inputs=[
|
198 |
+
prompt,
|
199 |
+
negative_prompt,
|
200 |
+
use_negative_prompt,
|
201 |
+
seed,
|
202 |
+
width,
|
203 |
+
height,
|
204 |
+
guidance_scale,
|
205 |
+
num_inference_steps,
|
206 |
+
randomize_seed,
|
207 |
+
],
|
208 |
+
outputs=[result, seed],
|
209 |
+
api_name="run",
|
210 |
)
|
211 |
|
212 |
+
if __name__ == "__main__":
|
213 |
+
demo.queue(max_size=20).launch()
|