Spaces:
Running
Running
Jinglong Xiong
commited on
Commit
·
bb5a422
1
Parent(s):
33dda45
can generate multiple variations
Browse files- gen_image.py +40 -7
gen_image.py
CHANGED
@@ -3,6 +3,15 @@ import torch
|
|
3 |
|
4 |
class ImageGenerator:
|
5 |
def __init__(self, model_id="stabilityai/stable-diffusion-2-1-base", device="cuda"):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
scheduler = EulerDiscreteScheduler.from_pretrained(model_id, subfolder="scheduler")
|
7 |
self.pipe = StableDiffusionPipeline.from_pretrained(
|
8 |
model_id,
|
@@ -10,25 +19,49 @@ class ImageGenerator:
|
|
10 |
torch_dtype=torch.float16
|
11 |
)
|
12 |
self.pipe = self.pipe.to(device)
|
|
|
|
|
13 |
|
14 |
-
def generate(self, prompt, negative_prompt=None, output_path=None):
|
15 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
|
17 |
if output_path:
|
18 |
-
image
|
|
|
19 |
|
20 |
return image
|
21 |
|
22 |
-
|
23 |
# Example usage
|
24 |
if __name__ == "__main__":
|
25 |
generator = ImageGenerator()
|
26 |
import time
|
27 |
start_time = time.time()
|
28 |
image = generator.generate(
|
29 |
-
prompt="magenta trapezoids layered on a transluscent silver sheet
|
30 |
-
|
31 |
-
|
32 |
)
|
33 |
end_time = time.time()
|
34 |
print(f"Time taken: {end_time - start_time} seconds")
|
|
|
3 |
|
4 |
class ImageGenerator:
|
5 |
def __init__(self, model_id="stabilityai/stable-diffusion-2-1-base", device="cuda"):
|
6 |
+
"""
|
7 |
+
Initialize the image generator with a specific model.
|
8 |
+
|
9 |
+
Args:
|
10 |
+
model_id (str): The model identifier for the stable diffusion model.
|
11 |
+
Default is "stabilityai/stable-diffusion-2-1-base".
|
12 |
+
device (str): The device to run the model on, either "cuda" or "cpu".
|
13 |
+
Default is "cuda".
|
14 |
+
"""
|
15 |
scheduler = EulerDiscreteScheduler.from_pretrained(model_id, subfolder="scheduler")
|
16 |
self.pipe = StableDiffusionPipeline.from_pretrained(
|
17 |
model_id,
|
|
|
19 |
torch_dtype=torch.float16
|
20 |
)
|
21 |
self.pipe = self.pipe.to(device)
|
22 |
+
self.positive_prompt = "simple, icon"
|
23 |
+
self.negative_prompt = "3d, blurry, complex geometry, realistic"
|
24 |
|
25 |
+
def generate(self, prompt, negative_prompt=None, output_path=None, num_images=1, num_inference_steps=50):
|
26 |
+
"""
|
27 |
+
Generate an image based on the provided prompt.
|
28 |
+
|
29 |
+
Args:
|
30 |
+
prompt (str): The text description to generate an image from.
|
31 |
+
negative_prompt (str, optional): Elements to avoid in the generated image.
|
32 |
+
If None, uses the default negative prompt.
|
33 |
+
output_path (str, optional): Path to save the generated image.
|
34 |
+
If None, the image is not saved to disk.
|
35 |
+
num_images (int, optional): Number of images to generate.
|
36 |
+
|
37 |
+
Returns:
|
38 |
+
PIL.Image.Image: The generated image.
|
39 |
+
"""
|
40 |
+
prompt = f"{prompt}, {self.positive_prompt}"
|
41 |
+
if negative_prompt is None:
|
42 |
+
negative_prompt = self.negative_prompt
|
43 |
+
images = self.pipe(
|
44 |
+
prompt,
|
45 |
+
negative_prompt=negative_prompt,
|
46 |
+
num_inference_steps=50,
|
47 |
+
num_images_per_prompt=num_images
|
48 |
+
).images
|
49 |
|
50 |
if output_path:
|
51 |
+
for i, image in enumerate(images):
|
52 |
+
image.save(f".cache/{output_path.replace('.png', f'_{i}.png')}")
|
53 |
|
54 |
return image
|
55 |
|
|
|
56 |
# Example usage
|
57 |
if __name__ == "__main__":
|
58 |
generator = ImageGenerator()
|
59 |
import time
|
60 |
start_time = time.time()
|
61 |
image = generator.generate(
|
62 |
+
prompt="magenta trapezoids layered on a transluscent silver sheet",
|
63 |
+
output_path="sheet.png",
|
64 |
+
num_images=4
|
65 |
)
|
66 |
end_time = time.time()
|
67 |
print(f"Time taken: {end_time - start_time} seconds")
|