Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -10,30 +10,31 @@ flash_pipe.scheduler = EulerDiscreteScheduler.from_config(flash_pipe.scheduler.c
|
|
10 |
clip_slider = CLIPSliderXL(flash_pipe, device=torch.device("cuda"), iterations=50)
|
11 |
|
12 |
@spaces.GPU
|
13 |
-
def generate(slider_x, slider_y, prompt, x_concept_1, x_concept_2, y_concept_1, y_concept_2):
|
14 |
|
15 |
# check if avg diff for directions need to be re-calculated
|
16 |
if not sorted(slider_x) == sorted([x_concept_1, x_concept_2]):
|
17 |
clip_slider.avg_diff = clip_slider.find_latent_direction(slider_x[0], slider_x[1])
|
18 |
x_concept_1, x_concept_2 = slider_x[0], slider_x[1]
|
19 |
-
|
20 |
-
print("clip_slider.avg_diff[1]", clip_slider.avg_diff[1])
|
21 |
if not sorted(slider_y) == sorted([y_concept_1, y_concept_2]):
|
22 |
clip_slider.avg_diff_2nd = clip_slider.find_latent_direction(slider_y[0], slider_y[1])
|
23 |
y_concept_1, y_concept_2 = slider_y[0], slider_y[1]
|
24 |
|
|
|
25 |
comma_concepts_x = ', '.join(slider_x)
|
26 |
comma_concepts_y = ', '.join(slider_y)
|
27 |
|
28 |
-
|
|
|
29 |
|
30 |
-
return gr.update(label=comma_concepts_x, interactive=True),gr.update(label=comma_concepts_y, interactive=True), x_concept_1, x_concept_2, y_concept_1, y_concept_2, image
|
31 |
|
32 |
-
def update_x(x,y,prompt):
|
33 |
image = clip_slider.generate(prompt, scale=x, scale_2nd=y, num_inference_steps=8)
|
34 |
return image
|
35 |
|
36 |
-
def update_y(x,y,prompt):
|
37 |
image = clip_slider.generate(prompt, scale=x, scale_2nd=y, num_inference_steps=8)
|
38 |
return image
|
39 |
|
@@ -67,6 +68,9 @@ with gr.Blocks(css=css) as demo:
|
|
67 |
x_concept_2 = gr.State("")
|
68 |
y_concept_1 = gr.State("")
|
69 |
y_concept_2 = gr.State("")
|
|
|
|
|
|
|
70 |
|
71 |
with gr.Row():
|
72 |
with gr.Column():
|
@@ -80,10 +84,10 @@ with gr.Blocks(css=css) as demo:
|
|
80 |
output_image = gr.Image(elem_id="image_out")
|
81 |
|
82 |
submit.click(fn=generate,
|
83 |
-
inputs=[slider_x, slider_y, prompt, x_concept_1, x_concept_2, y_concept_1, y_concept_2],
|
84 |
-
outputs=[x, y, x_concept_1, x_concept_2, y_concept_1, y_concept_2, output_image])
|
85 |
-
x.change(fn=update_x, inputs=[x,y, prompt], outputs=[output_image])
|
86 |
-
y.change(fn=update_y, inputs=[x,y, prompt], outputs=[output_image])
|
87 |
|
88 |
if __name__ == "__main__":
|
89 |
demo.launch()
|
|
|
10 |
clip_slider = CLIPSliderXL(flash_pipe, device=torch.device("cuda"), iterations=50)
|
11 |
|
12 |
@spaces.GPU
|
13 |
+
def generate(slider_x, slider_y, prompt, x_concept_1, x_concept_2, y_concept_1, y_concept_2, , avg_diff_x, avg_diff_y):
|
14 |
|
15 |
# check if avg diff for directions need to be re-calculated
|
16 |
if not sorted(slider_x) == sorted([x_concept_1, x_concept_2]):
|
17 |
clip_slider.avg_diff = clip_slider.find_latent_direction(slider_x[0], slider_x[1])
|
18 |
x_concept_1, x_concept_2 = slider_x[0], slider_x[1]
|
19 |
+
|
|
|
20 |
if not sorted(slider_y) == sorted([y_concept_1, y_concept_2]):
|
21 |
clip_slider.avg_diff_2nd = clip_slider.find_latent_direction(slider_y[0], slider_y[1])
|
22 |
y_concept_1, y_concept_2 = slider_y[0], slider_y[1]
|
23 |
|
24 |
+
image = clip_slider.generate(prompt, scale=0, scale_2nd=0, num_inference_steps=8)
|
25 |
comma_concepts_x = ', '.join(slider_x)
|
26 |
comma_concepts_y = ', '.join(slider_y)
|
27 |
|
28 |
+
avg_diff_x = clip_slider.avg_diff.cpu()
|
29 |
+
avg_diff_y = clip_slider.avg_diff_2nd.cpu()
|
30 |
|
31 |
+
return gr.update(label=comma_concepts_x, interactive=True),gr.update(label=comma_concepts_y, interactive=True), x_concept_1, x_concept_2, y_concept_1, y_concept_2, avg_diff_x, avg_diff_y, image
|
32 |
|
33 |
+
def update_x(x,y,prompt, avg_diff_x, avg_diff_y):
|
34 |
image = clip_slider.generate(prompt, scale=x, scale_2nd=y, num_inference_steps=8)
|
35 |
return image
|
36 |
|
37 |
+
def update_y(x,y,prompt, avg_diff_x, avg_diff_y):
|
38 |
image = clip_slider.generate(prompt, scale=x, scale_2nd=y, num_inference_steps=8)
|
39 |
return image
|
40 |
|
|
|
68 |
x_concept_2 = gr.State("")
|
69 |
y_concept_1 = gr.State("")
|
70 |
y_concept_2 = gr.State("")
|
71 |
+
|
72 |
+
avg_diff_x = gr.State()
|
73 |
+
avg_diff_y = gr.State()
|
74 |
|
75 |
with gr.Row():
|
76 |
with gr.Column():
|
|
|
84 |
output_image = gr.Image(elem_id="image_out")
|
85 |
|
86 |
submit.click(fn=generate,
|
87 |
+
inputs=[slider_x, slider_y, prompt, x_concept_1, x_concept_2, y_concept_1, y_concept_2, avg_diff_x, avg_diff_y],
|
88 |
+
outputs=[x, y, x_concept_1, x_concept_2, y_concept_1, y_concept_2, avg_diff_x, avg_diff_y, output_image])
|
89 |
+
x.change(fn=update_x, inputs=[x,y, prompt, avg_diff_x, avg_diff_y], outputs=[output_image])
|
90 |
+
y.change(fn=update_y, inputs=[x,y, prompt, avg_diff_x, avg_diff_y], outputs=[output_image])
|
91 |
|
92 |
if __name__ == "__main__":
|
93 |
demo.launch()
|