Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,157 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import torch
|
3 |
+
from diffusers import StableDiffusionXLPipeline, AutoencoderKL
|
4 |
+
from blora_utils import BLOCKS, filter_lora, scale_lora
|
5 |
+
|
6 |
+
vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
|
7 |
+
pipeline = StableDiffusionXLPipeline.from_pretrained(
|
8 |
+
"stabilityai/stable-diffusion-xl-base-1.0",
|
9 |
+
vae=vae,
|
10 |
+
torch_dtype=torch.float16,
|
11 |
+
).to("cuda")
|
12 |
+
|
13 |
+
def load_b_lora_to_unet(pipe, content_lora_model_id: str = '', style_lora_model_id: str = '', content_alpha: float = 1.,
|
14 |
+
style_alpha: float = 1.) -> None:
|
15 |
+
try:
|
16 |
+
# Get Content B-LoRA SD
|
17 |
+
if content_lora_model_id:
|
18 |
+
content_B_LoRA_sd, _ = pipe.lora_state_dict(content_lora_model_id)
|
19 |
+
content_B_LoRA = filter_lora(content_B_LoRA_sd, BLOCKS['content'])
|
20 |
+
content_B_LoRA = scale_lora(content_B_LoRA, content_alpha)
|
21 |
+
else:
|
22 |
+
content_B_LoRA = {}
|
23 |
+
|
24 |
+
# Get Style B-LoRA SD
|
25 |
+
if style_lora_model_id:
|
26 |
+
style_B_LoRA_sd, _ = pipe.lora_state_dict(style_lora_model_id)
|
27 |
+
style_B_LoRA = filter_lora(style_B_LoRA_sd, BLOCKS['style'])
|
28 |
+
style_B_LoRA = scale_lora(style_B_LoRA, style_alpha)
|
29 |
+
else:
|
30 |
+
style_B_LoRA = {}
|
31 |
+
|
32 |
+
# Merge B-LoRAs SD
|
33 |
+
res_lora = {**content_B_LoRA, **style_B_LoRA}
|
34 |
+
|
35 |
+
# Load
|
36 |
+
pipe.load_lora_into_unet(res_lora, None, pipe.unet)
|
37 |
+
except Exception as e:
|
38 |
+
raise type(e)(f'failed to load_b_lora_to_unet, due to: {e}')
|
39 |
+
|
40 |
+
def main(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps):
|
41 |
+
content_B_LoRA_path = ''
|
42 |
+
style_B_LoRA_path = 'fffiloni/b_lora_trained_test_7'
|
43 |
+
content_alpha,style_alpha = 1,1.1
|
44 |
+
|
45 |
+
load_b_lora_to_unet(pipeline, content_B_LoRA_path, style_B_LoRA_path, content_alpha, style_alpha)
|
46 |
+
prompt = 'An eagle in [v42] style'
|
47 |
+
image = pipeline(
|
48 |
+
prompt,
|
49 |
+
generator=torch.Generator(device="cuda").manual_seed(48),
|
50 |
+
num_images_per_prompt=1
|
51 |
+
).images[0].resize((512,512))
|
52 |
+
|
53 |
+
pipeline.unload_lora_weights()
|
54 |
+
|
55 |
+
return image
|
56 |
+
|
57 |
+
css="""
|
58 |
+
#col-container {
|
59 |
+
margin: 0 auto;
|
60 |
+
max-width: 520px;
|
61 |
+
}
|
62 |
+
"""
|
63 |
+
|
64 |
+
if torch.cuda.is_available():
|
65 |
+
power_device = "GPU"
|
66 |
+
else:
|
67 |
+
power_device = "CPU"
|
68 |
+
|
69 |
+
with gr.Blocks(css=css) as demo:
|
70 |
+
|
71 |
+
with gr.Column(elem_id="col-container"):
|
72 |
+
gr.Markdown(f"""
|
73 |
+
# Text-to-Image Gradio Template
|
74 |
+
Currently running on {power_device}.
|
75 |
+
""")
|
76 |
+
|
77 |
+
with gr.Row():
|
78 |
+
|
79 |
+
prompt = gr.Text(
|
80 |
+
label="Prompt",
|
81 |
+
show_label=False,
|
82 |
+
max_lines=1,
|
83 |
+
placeholder="Enter your prompt",
|
84 |
+
container=False,
|
85 |
+
)
|
86 |
+
|
87 |
+
run_button = gr.Button("Run", scale=0)
|
88 |
+
|
89 |
+
result = gr.Image(label="Result", show_label=False)
|
90 |
+
|
91 |
+
with gr.Accordion("Advanced Settings", open=False):
|
92 |
+
|
93 |
+
negative_prompt = gr.Text(
|
94 |
+
label="Negative prompt",
|
95 |
+
max_lines=1,
|
96 |
+
placeholder="Enter a negative prompt",
|
97 |
+
visible=False,
|
98 |
+
)
|
99 |
+
|
100 |
+
seed = gr.Slider(
|
101 |
+
label="Seed",
|
102 |
+
minimum=0,
|
103 |
+
maximum=MAX_SEED,
|
104 |
+
step=1,
|
105 |
+
value=0,
|
106 |
+
)
|
107 |
+
|
108 |
+
randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
|
109 |
+
|
110 |
+
with gr.Row():
|
111 |
+
|
112 |
+
width = gr.Slider(
|
113 |
+
label="Width",
|
114 |
+
minimum=256,
|
115 |
+
maximum=MAX_IMAGE_SIZE,
|
116 |
+
step=32,
|
117 |
+
value=512,
|
118 |
+
)
|
119 |
+
|
120 |
+
height = gr.Slider(
|
121 |
+
label="Height",
|
122 |
+
minimum=256,
|
123 |
+
maximum=MAX_IMAGE_SIZE,
|
124 |
+
step=32,
|
125 |
+
value=512,
|
126 |
+
)
|
127 |
+
|
128 |
+
with gr.Row():
|
129 |
+
|
130 |
+
guidance_scale = gr.Slider(
|
131 |
+
label="Guidance scale",
|
132 |
+
minimum=0.0,
|
133 |
+
maximum=10.0,
|
134 |
+
step=0.1,
|
135 |
+
value=0.0,
|
136 |
+
)
|
137 |
+
|
138 |
+
num_inference_steps = gr.Slider(
|
139 |
+
label="Number of inference steps",
|
140 |
+
minimum=1,
|
141 |
+
maximum=50,
|
142 |
+
step=1,
|
143 |
+
value=50,
|
144 |
+
)
|
145 |
+
|
146 |
+
gr.Examples(
|
147 |
+
examples = examples,
|
148 |
+
inputs = [prompt]
|
149 |
+
)
|
150 |
+
|
151 |
+
run_button.click(
|
152 |
+
fn = main,
|
153 |
+
inputs = [prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
|
154 |
+
outputs = [result]
|
155 |
+
)
|
156 |
+
|
157 |
+
demo.queue().launch()
|