Spaces:
Running
on
T4
Running
on
T4
Cache examples
Browse files
app.py
CHANGED
@@ -16,8 +16,12 @@ args.print_step = None
|
|
16 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
17 |
model, _, preprocess = open_clip.create_model_and_transforms(args.clip_model, pretrained=args.clip_pretrain, device=device)
|
18 |
|
|
|
19 |
|
20 |
def inference(target_image, prompt_len, iter):
|
|
|
|
|
|
|
21 |
if prompt_len is not None:
|
22 |
args.prompt_len = int(prompt_len)
|
23 |
else:
|
@@ -29,10 +33,13 @@ def inference(target_image, prompt_len, iter):
|
|
29 |
args.iter = 1000
|
30 |
|
31 |
learned_prompt = optimize_prompt(model, preprocess, args, device, target_images=[target_image])
|
32 |
-
|
33 |
return learned_prompt
|
34 |
-
|
35 |
def inference_text(target_prompt, prompt_len, iter):
|
|
|
|
|
|
|
36 |
if prompt_len is not None:
|
37 |
args.prompt_len = min(int(prompt_len), 75)
|
38 |
else:
|
@@ -85,7 +92,7 @@ with demo:
|
|
85 |
image_button.click(inference, inputs=[input_image, prompt_len_field, num_step_field], outputs=output_prompt)
|
86 |
prompt_button.click(inference_text, inputs=[input_prompt, prompt_len_field, num_step_field], outputs=output_prompt)
|
87 |
|
88 |
-
gr.Examples([["sample.jpeg", 8, 1000]], inputs=[input_image, prompt_len_field, num_step_field])
|
89 |
-
gr.Examples([["digital concept art of old wooden cabin in florida swamp, trending on artstation", 3, 1000]], inputs=[input_prompt, prompt_len_field, num_step_field])
|
90 |
|
91 |
demo.launch(enable_queue=True)
|
|
|
16 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
17 |
model, _, preprocess = open_clip.create_model_and_transforms(args.clip_model, pretrained=args.clip_pretrain, device=device)
|
18 |
|
19 |
+
args.counter = 0
|
20 |
|
21 |
def inference(target_image, prompt_len, iter):
|
22 |
+
args.counter += 1
|
23 |
+
print(args.counter)
|
24 |
+
|
25 |
if prompt_len is not None:
|
26 |
args.prompt_len = int(prompt_len)
|
27 |
else:
|
|
|
33 |
args.iter = 1000
|
34 |
|
35 |
learned_prompt = optimize_prompt(model, preprocess, args, device, target_images=[target_image])
|
36 |
+
|
37 |
return learned_prompt
|
38 |
+
|
39 |
def inference_text(target_prompt, prompt_len, iter):
|
40 |
+
args.counter += 1
|
41 |
+
print(args.counter)
|
42 |
+
|
43 |
if prompt_len is not None:
|
44 |
args.prompt_len = min(int(prompt_len), 75)
|
45 |
else:
|
|
|
92 |
image_button.click(inference, inputs=[input_image, prompt_len_field, num_step_field], outputs=output_prompt)
|
93 |
prompt_button.click(inference_text, inputs=[input_prompt, prompt_len_field, num_step_field], outputs=output_prompt)
|
94 |
|
95 |
+
gr.Examples([["sample.jpeg", 8, 1000]], inputs=[input_image, prompt_len_field, num_step_field], fn=inference, outputs=output_prompt, cache_examples=True)
|
96 |
+
gr.Examples([["digital concept art of old wooden cabin in florida swamp, trending on artstation", 3, 1000]], inputs=[input_prompt, prompt_len_field, num_step_field], fn=inference_text, outputs=output_prompt, cache_examples=True)
|
97 |
|
98 |
demo.launch(enable_queue=True)
|