Spaces:
Running
on
Zero
Running
on
Zero
Upload folder using huggingface_hub
Browse files
app.py
CHANGED
@@ -40,6 +40,7 @@ def inference(content_image, style_image, style_strength, output_quality, progre
|
|
40 |
print('-'*15)
|
41 |
print('DATETIME:', datetime.datetime.now())
|
42 |
print('STYLE:', style_image)
|
|
|
43 |
img_size = 1024 if output_quality else 512
|
44 |
content_img, original_size = preprocess_img(content_image, img_size)
|
45 |
content_img = content_img.to(device)
|
@@ -58,10 +59,14 @@ def inference(content_image, style_image, style_strength, output_quality, progre
|
|
58 |
st = time.time()
|
59 |
generated_img = content_img.clone().requires_grad_(True)
|
60 |
optimizer = optim.Adam([generated_img], lr=lr)
|
61 |
-
|
62 |
-
|
63 |
content_features = model(content_img)
|
64 |
style_features = model(style_img)
|
|
|
|
|
|
|
|
|
65 |
generated_features = model(generated_img)
|
66 |
|
67 |
content_loss = 0
|
@@ -81,7 +86,6 @@ def inference(content_image, style_image, style_strength, output_quality, progre
|
|
81 |
|
82 |
total_loss = alpha * content_loss + beta * style_loss
|
83 |
|
84 |
-
optimizer.zero_grad()
|
85 |
total_loss.backward()
|
86 |
optimizer.step()
|
87 |
|
|
|
40 |
print('-'*15)
|
41 |
print('DATETIME:', datetime.datetime.now())
|
42 |
print('STYLE:', style_image)
|
43 |
+
|
44 |
img_size = 1024 if output_quality else 512
|
45 |
content_img, original_size = preprocess_img(content_image, img_size)
|
46 |
content_img = content_img.to(device)
|
|
|
59 |
st = time.time()
|
60 |
generated_img = content_img.clone().requires_grad_(True)
|
61 |
optimizer = optim.Adam([generated_img], lr=lr)
|
62 |
+
|
63 |
+
with torch.no_grad():
|
64 |
content_features = model(content_img)
|
65 |
style_features = model(style_img)
|
66 |
+
|
67 |
+
for _ in tqdm(range(iters), desc='The magic is happening ✨'):
|
68 |
+
optimizer.zero_grad()
|
69 |
+
|
70 |
generated_features = model(generated_img)
|
71 |
|
72 |
content_loss = 0
|
|
|
86 |
|
87 |
total_loss = alpha * content_loss + beta * style_loss
|
88 |
|
|
|
89 |
total_loss.backward()
|
90 |
optimizer.step()
|
91 |
|