Spaces:
Runtime error
Runtime error
RamAnanth1
commited on
Commit
•
02fa9c7
1
Parent(s):
6585503
Update app.py
Browse files
app.py
CHANGED
@@ -20,13 +20,6 @@ canny_checkpoint = "models/control_sd15_canny.pth"
|
|
20 |
scribble_checkpoint = "models/control_sd15_scribble.pth"
|
21 |
pose_checkpoint = "models/control_sd15_openpose.pth"
|
22 |
|
23 |
-
canny_model = create_model('./models/cldm_v15.yaml').cpu()
|
24 |
-
canny_model.load_state_dict(load_state_dict(cached_download(
|
25 |
-
hf_hub_url(REPO_ID, canny_checkpoint)
|
26 |
-
), location='cuda'))
|
27 |
-
canny_model = canny_model.cuda()
|
28 |
-
ddim_sampler = DDIMSampler(canny_model)
|
29 |
-
|
30 |
pose_model = create_model('./models/cldm_v15.yaml').cpu()
|
31 |
pose_model.load_state_dict(load_state_dict(cached_download(
|
32 |
hf_hub_url(REPO_ID, pose_checkpoint)
|
@@ -34,10 +27,10 @@ pose_model.load_state_dict(load_state_dict(cached_download(
|
|
34 |
pose_model = pose_model.cuda()
|
35 |
ddim_sampler_pose = DDIMSampler(pose_model)
|
36 |
|
37 |
-
scribble_model = create_model('./models/cldm_v15.yaml')
|
38 |
scribble_model.load_state_dict(load_state_dict(cached_download(
|
39 |
hf_hub_url(REPO_ID, scribble_checkpoint)
|
40 |
-
), location='
|
41 |
scribble_model = canny_model.cuda()
|
42 |
ddim_sampler_scribble = DDIMSampler(scribble_model)
|
43 |
|
@@ -45,38 +38,9 @@ def process(input_image, prompt, input_control, a_prompt, n_prompt, num_samples,
|
|
45 |
# TODO: Add other control tasks
|
46 |
if input_control == "Scribble":
|
47 |
return process_scribble(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta)
|
48 |
-
|
49 |
return process_pose(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, image_resolution, ddim_steps, scale, seed, eta)
|
50 |
-
return process_canny(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta, low_threshold, high_threshold)
|
51 |
|
52 |
-
def process_canny(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta, low_threshold, high_threshold):
|
53 |
-
with torch.no_grad():
|
54 |
-
img = resize_image(HWC3(input_image), image_resolution)
|
55 |
-
H, W, C = img.shape
|
56 |
-
|
57 |
-
detected_map = apply_canny(img, low_threshold, high_threshold)
|
58 |
-
detected_map = HWC3(detected_map)
|
59 |
-
|
60 |
-
control = torch.from_numpy(detected_map.copy()).float().cuda() / 255.0
|
61 |
-
control = torch.stack([control for _ in range(num_samples)], dim=0)
|
62 |
-
control = einops.rearrange(control, 'b h w c -> b c h w').clone()
|
63 |
-
|
64 |
-
seed_everything(seed)
|
65 |
-
|
66 |
-
cond = {"c_concat": [control], "c_crossattn": [canny_model.get_learned_conditioning([prompt + ', ' + a_prompt] * num_samples)]}
|
67 |
-
un_cond = {"c_concat": [control], "c_crossattn": [canny_model.get_learned_conditioning([n_prompt] * num_samples)]}
|
68 |
-
shape = (4, H // 8, W // 8)
|
69 |
-
|
70 |
-
samples, intermediates = ddim_sampler.sample(ddim_steps, num_samples,
|
71 |
-
shape, cond, verbose=False, eta=eta,
|
72 |
-
unconditional_guidance_scale=scale,
|
73 |
-
unconditional_conditioning=un_cond)
|
74 |
-
x_samples = canny_model.decode_first_stage(samples)
|
75 |
-
x_samples = (einops.rearrange(x_samples, 'b c h w -> b h w c') * 127.5 + 127.5).cpu().numpy().clip(0, 255).astype(np.uint8)
|
76 |
-
|
77 |
-
results = [x_samples[i] for i in range(num_samples)]
|
78 |
-
return [255 - detected_map] + results
|
79 |
-
|
80 |
def process_scribble(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta):
|
81 |
with torch.no_grad():
|
82 |
img = resize_image(HWC3(input_image), image_resolution)
|
@@ -150,7 +114,6 @@ def create_canvas(w, h):
|
|
150 |
|
151 |
block = gr.Blocks().queue()
|
152 |
control_task_list = [
|
153 |
-
"Canny Edge Map",
|
154 |
"Scribble",
|
155 |
"Pose"
|
156 |
]
|
@@ -172,8 +135,6 @@ with block:
|
|
172 |
with gr.Accordion("Advanced options", open=False):
|
173 |
num_samples = gr.Slider(label="Images", minimum=1, maximum=12, value=1, step=1)
|
174 |
image_resolution = gr.Slider(label="Image Resolution", minimum=256, maximum=768, value=512, step=256)
|
175 |
-
low_threshold = gr.Slider(label="Canny low threshold", minimum=1, maximum=255, value=100, step=1)
|
176 |
-
high_threshold = gr.Slider(label="Canny high threshold", minimum=1, maximum=255, value=200, step=1)
|
177 |
ddim_steps = gr.Slider(label="Steps", minimum=1, maximum=100, value=20, step=1)
|
178 |
scale = gr.Slider(label="Guidance Scale", minimum=0.1, maximum=30.0, value=9.0, step=0.1)
|
179 |
seed = gr.Slider(label="Seed", minimum=0, maximum=2147483647, step=1, randomize=True)
|
@@ -183,25 +144,10 @@ with block:
|
|
183 |
value='longbody, lowres, bad anatomy, bad hands, missing fingers, pubic hair,extra digit, fewer digits, cropped, worst quality, low quality')
|
184 |
with gr.Column():
|
185 |
result_gallery = gr.Gallery(label='Output', show_label=False, elem_id="gallery").style(grid=2, height='auto')
|
186 |
-
ips = [input_image, prompt, input_control, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta
|
187 |
run_button.click(fn=process, inputs=ips, outputs=[result_gallery])
|
188 |
examples_list = [
|
189 |
-
|
190 |
-
"bird.png",
|
191 |
-
"bird",
|
192 |
-
"Canny Edge Map",
|
193 |
-
"best quality, extremely detailed",
|
194 |
-
'longbody, lowres, bad anatomy, bad hands, missing fingers, pubic hair,extra digit, fewer digits, cropped, worst quality, low quality',
|
195 |
-
1,
|
196 |
-
512,
|
197 |
-
20,
|
198 |
-
9.0,
|
199 |
-
123490213,
|
200 |
-
0.0,
|
201 |
-
100,
|
202 |
-
200
|
203 |
-
|
204 |
-
],
|
205 |
[
|
206 |
"turtle.png",
|
207 |
"turtle",
|
@@ -213,9 +159,7 @@ with block:
|
|
213 |
20,
|
214 |
9.0,
|
215 |
123490213,
|
216 |
-
0.0
|
217 |
-
100,
|
218 |
-
200
|
219 |
|
220 |
]
|
221 |
]
|
|
|
20 |
scribble_checkpoint = "models/control_sd15_scribble.pth"
|
21 |
pose_checkpoint = "models/control_sd15_openpose.pth"
|
22 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
pose_model = create_model('./models/cldm_v15.yaml').cpu()
|
24 |
pose_model.load_state_dict(load_state_dict(cached_download(
|
25 |
hf_hub_url(REPO_ID, pose_checkpoint)
|
|
|
27 |
pose_model = pose_model.cuda()
|
28 |
ddim_sampler_pose = DDIMSampler(pose_model)
|
29 |
|
30 |
+
scribble_model = create_model('./models/cldm_v15.yaml').cpu()
|
31 |
scribble_model.load_state_dict(load_state_dict(cached_download(
|
32 |
hf_hub_url(REPO_ID, scribble_checkpoint)
|
33 |
+
), location='cuda'))
|
34 |
scribble_model = canny_model.cuda()
|
35 |
ddim_sampler_scribble = DDIMSampler(scribble_model)
|
36 |
|
|
|
38 |
# TODO: Add other control tasks
|
39 |
if input_control == "Scribble":
|
40 |
return process_scribble(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta)
|
41 |
+
else:
|
42 |
return process_pose(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, image_resolution, ddim_steps, scale, seed, eta)
|
|
|
43 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
44 |
def process_scribble(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta):
|
45 |
with torch.no_grad():
|
46 |
img = resize_image(HWC3(input_image), image_resolution)
|
|
|
114 |
|
115 |
block = gr.Blocks().queue()
|
116 |
control_task_list = [
|
|
|
117 |
"Scribble",
|
118 |
"Pose"
|
119 |
]
|
|
|
135 |
with gr.Accordion("Advanced options", open=False):
|
136 |
num_samples = gr.Slider(label="Images", minimum=1, maximum=12, value=1, step=1)
|
137 |
image_resolution = gr.Slider(label="Image Resolution", minimum=256, maximum=768, value=512, step=256)
|
|
|
|
|
138 |
ddim_steps = gr.Slider(label="Steps", minimum=1, maximum=100, value=20, step=1)
|
139 |
scale = gr.Slider(label="Guidance Scale", minimum=0.1, maximum=30.0, value=9.0, step=0.1)
|
140 |
seed = gr.Slider(label="Seed", minimum=0, maximum=2147483647, step=1, randomize=True)
|
|
|
144 |
value='longbody, lowres, bad anatomy, bad hands, missing fingers, pubic hair,extra digit, fewer digits, cropped, worst quality, low quality')
|
145 |
with gr.Column():
|
146 |
result_gallery = gr.Gallery(label='Output', show_label=False, elem_id="gallery").style(grid=2, height='auto')
|
147 |
+
ips = [input_image, prompt, input_control, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta]
|
148 |
run_button.click(fn=process, inputs=ips, outputs=[result_gallery])
|
149 |
examples_list = [
|
150 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
151 |
[
|
152 |
"turtle.png",
|
153 |
"turtle",
|
|
|
159 |
20,
|
160 |
9.0,
|
161 |
123490213,
|
162 |
+
0.0
|
|
|
|
|
163 |
|
164 |
]
|
165 |
]
|