Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -49,14 +49,17 @@ def initialize_models():
|
|
49 |
# Load pipeline
|
50 |
pipeline = load_pipeline(args, accelerator, enable_xformers_memory_efficient_attention=False)
|
51 |
|
52 |
-
#
|
53 |
pipeline.unet.eval()
|
54 |
pipeline.controlnet.eval()
|
55 |
pipeline.vae.eval()
|
56 |
pipeline.text_encoder.eval()
|
57 |
|
|
|
|
|
|
|
58 |
# Initialize generator
|
59 |
-
generator = torch.Generator(
|
60 |
|
61 |
return True
|
62 |
|
@@ -64,7 +67,7 @@ def initialize_models():
|
|
64 |
print(f"Error initializing models: {str(e)}")
|
65 |
return False
|
66 |
|
67 |
-
@spaces.GPU(processing_timeout=180)
|
68 |
def process_image(
|
69 |
input_image,
|
70 |
prompt="clean, high-resolution, 8k",
|
@@ -78,11 +81,12 @@ def process_image(
|
|
78 |
):
|
79 |
global pipeline, generator, accelerator
|
80 |
|
81 |
-
if pipeline is None:
|
82 |
-
if not initialize_models():
|
83 |
-
return None
|
84 |
-
|
85 |
try:
|
|
|
|
|
|
|
|
|
|
|
86 |
# Create args object with all necessary parameters
|
87 |
args = Args(
|
88 |
added_prompt=prompt,
|
@@ -124,27 +128,38 @@ def process_image(
|
|
124 |
validation_image = validation_image.resize((validation_image.size[0]//8*8, validation_image.size[1]//8*8))
|
125 |
width, height = validation_image.size
|
126 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
127 |
# Generate image
|
128 |
with torch.no_grad():
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
|
136 |
-
|
137 |
-
|
138 |
-
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
-
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
-
|
|
|
|
|
|
|
|
|
148 |
|
149 |
image = output.images[0]
|
150 |
|
@@ -161,6 +176,8 @@ def process_image(
|
|
161 |
|
162 |
except Exception as e:
|
163 |
print(f"Error processing image: {str(e)}")
|
|
|
|
|
164 |
return None
|
165 |
|
166 |
# Create Gradio interface
|
|
|
49 |
# Load pipeline
|
50 |
pipeline = load_pipeline(args, accelerator, enable_xformers_memory_efficient_attention=False)
|
51 |
|
52 |
+
# Ensure all models are in eval mode
|
53 |
pipeline.unet.eval()
|
54 |
pipeline.controlnet.eval()
|
55 |
pipeline.vae.eval()
|
56 |
pipeline.text_encoder.eval()
|
57 |
|
58 |
+
# Move pipeline to CUDA
|
59 |
+
pipeline = pipeline.to("cuda")
|
60 |
+
|
61 |
# Initialize generator
|
62 |
+
generator = torch.Generator("cuda")
|
63 |
|
64 |
return True
|
65 |
|
|
|
67 |
print(f"Error initializing models: {str(e)}")
|
68 |
return False
|
69 |
|
70 |
+
@spaces.GPU(processing_timeout=180)
|
71 |
def process_image(
|
72 |
input_image,
|
73 |
prompt="clean, high-resolution, 8k",
|
|
|
81 |
):
|
82 |
global pipeline, generator, accelerator
|
83 |
|
|
|
|
|
|
|
|
|
84 |
try:
|
85 |
+
# Initialize models if not already done
|
86 |
+
if pipeline is None:
|
87 |
+
if not initialize_models():
|
88 |
+
return None
|
89 |
+
|
90 |
# Create args object with all necessary parameters
|
91 |
args = Args(
|
92 |
added_prompt=prompt,
|
|
|
128 |
validation_image = validation_image.resize((validation_image.size[0]//8*8, validation_image.size[1]//8*8))
|
129 |
width, height = validation_image.size
|
130 |
|
131 |
+
# Ensure pipeline is on CUDA and in eval mode
|
132 |
+
pipeline = pipeline.to("cuda")
|
133 |
+
pipeline.unet.eval()
|
134 |
+
pipeline.controlnet.eval()
|
135 |
+
pipeline.vae.eval()
|
136 |
+
pipeline.text_encoder.eval()
|
137 |
+
|
138 |
# Generate image
|
139 |
with torch.no_grad():
|
140 |
+
try:
|
141 |
+
inference_time, output = pipeline(
|
142 |
+
args.t_max,
|
143 |
+
args.t_min,
|
144 |
+
args.tile_diffusion,
|
145 |
+
args.tile_diffusion_size,
|
146 |
+
args.tile_diffusion_stride,
|
147 |
+
args.added_prompt,
|
148 |
+
validation_image,
|
149 |
+
num_inference_steps=args.num_inference_steps,
|
150 |
+
generator=generator,
|
151 |
+
height=height,
|
152 |
+
width=width,
|
153 |
+
guidance_scale=args.guidance_scale,
|
154 |
+
negative_prompt=args.negative_prompt,
|
155 |
+
conditioning_scale=args.conditioning_scale,
|
156 |
+
start_steps=args.start_steps,
|
157 |
+
start_point=args.start_point,
|
158 |
+
use_vae_encode_condition=args.use_vae_encode_condition,
|
159 |
+
)
|
160 |
+
except Exception as e:
|
161 |
+
print(f"Pipeline execution error: {str(e)}")
|
162 |
+
raise
|
163 |
|
164 |
image = output.images[0]
|
165 |
|
|
|
176 |
|
177 |
except Exception as e:
|
178 |
print(f"Error processing image: {str(e)}")
|
179 |
+
import traceback
|
180 |
+
traceback.print_exc()
|
181 |
return None
|
182 |
|
183 |
# Create Gradio interface
|