Spaces:
Running
on
Zero
Running
on
Zero
zhiweili
commited on
Commit
•
01b766d
1
Parent(s):
21ef723
change base model
Browse files- app_haircolor.py +30 -9
app_haircolor.py
CHANGED
@@ -11,13 +11,15 @@ from segment_utils import(
|
|
11 |
from diffusers import (
|
12 |
DiffusionPipeline,
|
13 |
T2IAdapter,
|
|
|
14 |
)
|
15 |
|
16 |
from controlnet_aux import (
|
17 |
LineartDetector,
|
|
|
18 |
)
|
19 |
|
20 |
-
BASE_MODEL = "stabilityai/
|
21 |
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
22 |
|
23 |
DEFAULT_EDIT_PROMPT = "a woman, blue hair, high detailed"
|
@@ -28,18 +30,30 @@ DEFAULT_CATEGORY = "hair"
|
|
28 |
lineart_detector = LineartDetector.from_pretrained("lllyasviel/Annotators")
|
29 |
lineart_detector = lineart_detector.to(DEVICE)
|
30 |
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
35 |
)
|
|
|
36 |
|
37 |
|
38 |
basepipeline = DiffusionPipeline.from_pretrained(
|
39 |
BASE_MODEL,
|
40 |
torch_dtype=torch.float16,
|
41 |
use_safetensors=True,
|
42 |
-
adapter=
|
43 |
custom_pipeline="./pipelines/pipeline_sdxl_adapter_img2img.py",
|
44 |
)
|
45 |
|
@@ -55,13 +69,20 @@ def image_to_image(
|
|
55 |
num_steps: int,
|
56 |
guidance_scale: float,
|
57 |
generate_size: int,
|
58 |
-
|
|
|
59 |
):
|
60 |
run_task_time = 0
|
61 |
time_cost_str = ''
|
62 |
run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
|
63 |
lineart_image = lineart_detector(input_image, int(generate_size*0.375), generate_size)
|
64 |
run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
|
|
|
|
|
|
|
|
|
|
|
|
|
65 |
generator = torch.Generator(device=DEVICE).manual_seed(seed)
|
66 |
generated_image = basepipeline(
|
67 |
generator=generator,
|
@@ -72,8 +93,8 @@ def image_to_image(
|
|
72 |
width=generate_size,
|
73 |
guidance_scale=guidance_scale,
|
74 |
num_inference_steps=num_steps,
|
75 |
-
adapter_image=
|
76 |
-
adapter_conditioning_scale=
|
77 |
).images[0]
|
78 |
|
79 |
run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
|
|
|
11 |
from diffusers import (
|
12 |
DiffusionPipeline,
|
13 |
T2IAdapter,
|
14 |
+
MultiAdapter,
|
15 |
)
|
16 |
|
17 |
from controlnet_aux import (
|
18 |
LineartDetector,
|
19 |
+
CannyDetector,
|
20 |
)
|
21 |
|
22 |
+
BASE_MODEL = "stabilityai/stable-diffusion-xl-base-1.0"
|
23 |
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
24 |
|
25 |
DEFAULT_EDIT_PROMPT = "a woman, blue hair, high detailed"
|
|
|
30 |
lineart_detector = LineartDetector.from_pretrained("lllyasviel/Annotators")
|
31 |
lineart_detector = lineart_detector.to(DEVICE)
|
32 |
|
33 |
+
canndy_detector = CannyDetector()
|
34 |
+
|
35 |
+
adapters = MultiAdapter(
|
36 |
+
[
|
37 |
+
T2IAdapter.from_pretrained(
|
38 |
+
"TencentARC/t2i-adapter-lineart-sdxl-1.0",
|
39 |
+
torch_dtype=torch.float16,
|
40 |
+
varient="fp16",
|
41 |
+
),
|
42 |
+
T2IAdapter.from_pretrained(
|
43 |
+
"TencentARC/t2i-adapter-canny-sdxl-1.0",
|
44 |
+
torch_dtype=torch.float16,
|
45 |
+
varient="fp16",
|
46 |
+
),
|
47 |
+
]
|
48 |
)
|
49 |
+
adapters = adapters.to(torch.float16)
|
50 |
|
51 |
|
52 |
basepipeline = DiffusionPipeline.from_pretrained(
|
53 |
BASE_MODEL,
|
54 |
torch_dtype=torch.float16,
|
55 |
use_safetensors=True,
|
56 |
+
adapter=adapters,
|
57 |
custom_pipeline="./pipelines/pipeline_sdxl_adapter_img2img.py",
|
58 |
)
|
59 |
|
|
|
69 |
num_steps: int,
|
70 |
guidance_scale: float,
|
71 |
generate_size: int,
|
72 |
+
lineart_scale: float = 1.0,
|
73 |
+
canny_scale: float = 0.5,
|
74 |
):
|
75 |
run_task_time = 0
|
76 |
time_cost_str = ''
|
77 |
run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
|
78 |
lineart_image = lineart_detector(input_image, int(generate_size*0.375), generate_size)
|
79 |
run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
|
80 |
+
canny_image = canndy_detector(input_image, int(generate_size*0.375), generate_size)
|
81 |
+
run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
|
82 |
+
|
83 |
+
cond_image = [lineart_image, canny_image]
|
84 |
+
cond_scale = [lineart_scale, canny_scale]
|
85 |
+
|
86 |
generator = torch.Generator(device=DEVICE).manual_seed(seed)
|
87 |
generated_image = basepipeline(
|
88 |
generator=generator,
|
|
|
93 |
width=generate_size,
|
94 |
guidance_scale=guidance_scale,
|
95 |
num_inference_steps=num_steps,
|
96 |
+
adapter_image=cond_image,
|
97 |
+
adapter_conditioning_scale=cond_scale,
|
98 |
).images[0]
|
99 |
|
100 |
run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
|