Spaces:
Runtime error
Runtime error
Update flux1_img2img.py
Browse files- flux1_img2img.py +64 -80
flux1_img2img.py
CHANGED
@@ -1,6 +1,4 @@
|
|
1 |
import os
|
2 |
-
import re
|
3 |
-
import sys
|
4 |
import torch
|
5 |
import gradio as gr
|
6 |
from PIL import Image
|
@@ -9,20 +7,20 @@ import spaces
|
|
9 |
from diffusers import FluxImg2ImgPipeline
|
10 |
|
11 |
###############################################################################
|
12 |
-
#
|
13 |
###############################################################################
|
14 |
-
pipe = None # We
|
15 |
|
16 |
###############################################################################
|
17 |
-
#
|
18 |
###############################################################################
|
19 |
-
def resize_image(image,
|
20 |
"""
|
21 |
-
Resizes
|
22 |
-
|
23 |
"""
|
24 |
w, h = image.size
|
25 |
-
ratio = min(
|
26 |
if ratio < 1.0:
|
27 |
new_w = int(w * ratio)
|
28 |
new_h = int(h * ratio)
|
@@ -30,96 +28,93 @@ def resize_image(image, max_size=512):
|
|
30 |
return image
|
31 |
|
32 |
###############################################################################
|
33 |
-
#
|
34 |
###############################################################################
|
35 |
def load_flux_pipeline():
|
36 |
-
"""
|
37 |
-
Lazily loads the FluxImg2ImgPipeline with float16,
|
38 |
-
CPU offload, xFormers (if installed), etc.
|
39 |
-
"""
|
40 |
global pipe
|
41 |
if pipe is not None:
|
42 |
return # Already loaded
|
43 |
-
|
44 |
-
print("Loading
|
45 |
-
|
|
|
46 |
pipe_local = FluxImg2ImgPipeline.from_pretrained(
|
47 |
"black-forest-labs/FLUX.1-schnell",
|
48 |
-
torch_dtype=torch.float16,
|
49 |
low_cpu_mem_usage=True
|
50 |
)
|
51 |
|
52 |
-
# Move to GPU
|
53 |
pipe_local.to("cuda")
|
54 |
|
55 |
-
#
|
56 |
try:
|
57 |
pipe_local.enable_xformers_memory_efficient_attention()
|
58 |
-
print("
|
59 |
except Exception as e:
|
60 |
print("Could not enable xFormers:", e)
|
61 |
|
62 |
-
#
|
63 |
try:
|
64 |
pipe_local.enable_model_cpu_offload()
|
65 |
-
print("
|
66 |
except Exception as e:
|
67 |
print("Could not enable model_cpu_offload:", e)
|
68 |
|
69 |
-
# VAE slicing
|
70 |
pipe_local.enable_vae_slicing()
|
71 |
|
72 |
-
|
73 |
-
|
74 |
pipe = pipe_local
|
|
|
75 |
|
76 |
###############################################################################
|
77 |
-
#
|
78 |
###############################################################################
|
79 |
@spaces.GPU
|
80 |
def process_image(
|
81 |
-
image,
|
82 |
-
mask_image,
|
83 |
-
prompt="
|
84 |
-
model_id="black-forest-labs/FLUX.1-schnell",
|
85 |
strength=0.75,
|
86 |
seed=0,
|
87 |
num_inference_steps=4,
|
88 |
progress=gr.Progress(track_tqdm=True)
|
89 |
):
|
90 |
"""
|
91 |
-
|
92 |
-
|
93 |
"""
|
|
|
94 |
|
95 |
-
#
|
96 |
-
|
97 |
|
|
|
98 |
if image is None:
|
99 |
print("No input image provided.")
|
100 |
return None
|
101 |
|
102 |
-
#
|
103 |
-
|
104 |
-
|
105 |
-
# 2) Resize input to reduce VRAM usage
|
106 |
-
image = resize_image(image, max_size=512)
|
107 |
|
108 |
-
# 3)
|
109 |
generator = torch.Generator("cuda").manual_seed(seed)
|
110 |
-
|
111 |
-
# 4)
|
112 |
-
|
|
|
113 |
output = pipe(
|
114 |
prompt=prompt,
|
115 |
image=image,
|
116 |
generator=generator,
|
117 |
strength=strength,
|
118 |
-
guidance_scale=0,
|
119 |
num_inference_steps=num_inference_steps
|
120 |
)
|
121 |
-
progress(100, desc="Done")
|
122 |
|
|
|
123 |
return output.images[0]
|
124 |
|
125 |
###############################################################################
|
@@ -137,57 +132,46 @@ css = """
|
|
137 |
"""
|
138 |
|
139 |
with gr.Blocks(css=css) as demo:
|
140 |
-
gr.Markdown("##
|
|
|
141 |
|
142 |
with gr.Row():
|
143 |
with gr.Column():
|
144 |
-
|
|
|
145 |
label="Input Image (Img2Img)",
|
146 |
type="pil",
|
147 |
image_mode="RGB",
|
148 |
height=512
|
149 |
)
|
150 |
-
|
151 |
-
|
|
|
152 |
label="Mask (unused)",
|
153 |
type="pil",
|
154 |
image_mode="RGB",
|
155 |
-
height=
|
156 |
-
)
|
157 |
-
prompt_input = gr.Textbox(label="Prompt", value="a person")
|
158 |
-
strength_slider = gr.Slider(
|
159 |
-
minimum=0.0,
|
160 |
-
maximum=1.0,
|
161 |
-
value=0.75,
|
162 |
-
step=0.05,
|
163 |
-
label="Strength"
|
164 |
-
)
|
165 |
-
seed_box = gr.Number(label="Seed", value=0)
|
166 |
-
steps_box = gr.Slider(
|
167 |
-
minimum=1,
|
168 |
-
maximum=50,
|
169 |
-
value=4,
|
170 |
-
step=1,
|
171 |
-
label="Inference Steps"
|
172 |
)
|
173 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
174 |
|
175 |
with gr.Column():
|
176 |
-
|
|
|
|
|
|
|
|
|
177 |
|
178 |
-
#
|
179 |
run_button.click(
|
180 |
fn=process_image,
|
181 |
-
inputs=[
|
182 |
-
|
183 |
-
mask_input,
|
184 |
-
prompt_input,
|
185 |
-
# model_id is default, so we won't pass it from UI
|
186 |
-
strength_slider,
|
187 |
-
seed_box,
|
188 |
-
steps_box
|
189 |
-
],
|
190 |
-
outputs=[output_image]
|
191 |
)
|
192 |
|
193 |
if __name__ == "__main__":
|
|
|
1 |
import os
|
|
|
|
|
2 |
import torch
|
3 |
import gradio as gr
|
4 |
from PIL import Image
|
|
|
7 |
from diffusers import FluxImg2ImgPipeline
|
8 |
|
9 |
###############################################################################
|
10 |
+
# GLOBALS
|
11 |
###############################################################################
|
12 |
+
pipe = None # We'll load it lazily to avoid OOM during space startup
|
13 |
|
14 |
###############################################################################
|
15 |
+
# Helper: Resize the input image
|
16 |
###############################################################################
|
17 |
+
def resize_image(image: Image.Image, max_dim: int = 512) -> Image.Image:
|
18 |
"""
|
19 |
+
Resizes 'image' so that its largest dimension <= max_dim,
|
20 |
+
preserving aspect ratio. This helps reduce VRAM usage on T4.
|
21 |
"""
|
22 |
w, h = image.size
|
23 |
+
ratio = min(max_dim / w, max_dim / h)
|
24 |
if ratio < 1.0:
|
25 |
new_w = int(w * ratio)
|
26 |
new_h = int(h * ratio)
|
|
|
28 |
return image
|
29 |
|
30 |
###############################################################################
|
31 |
+
# Lazy-load function for FLUX.1-schnell pipeline in float16
|
32 |
###############################################################################
|
33 |
def load_flux_pipeline():
|
|
|
|
|
|
|
|
|
34 |
global pipe
|
35 |
if pipe is not None:
|
36 |
return # Already loaded
|
37 |
+
|
38 |
+
print("Loading FLUX.1-schnell with float16 on T4...")
|
39 |
+
|
40 |
+
# 1) Load in float16 (NOT bfloat16)
|
41 |
pipe_local = FluxImg2ImgPipeline.from_pretrained(
|
42 |
"black-forest-labs/FLUX.1-schnell",
|
43 |
+
torch_dtype=torch.float16, # crucial for T4
|
44 |
low_cpu_mem_usage=True
|
45 |
)
|
46 |
|
47 |
+
# 2) Move to GPU
|
48 |
pipe_local.to("cuda")
|
49 |
|
50 |
+
# 3) Memory Efficient Attention (xFormers)
|
51 |
try:
|
52 |
pipe_local.enable_xformers_memory_efficient_attention()
|
53 |
+
print("xFormers memory efficient attention enabled.")
|
54 |
except Exception as e:
|
55 |
print("Could not enable xFormers:", e)
|
56 |
|
57 |
+
# 4) CPU offload (keeps only active layers on GPU)
|
58 |
try:
|
59 |
pipe_local.enable_model_cpu_offload()
|
60 |
+
print("Model CPU offload enabled.")
|
61 |
except Exception as e:
|
62 |
print("Could not enable model_cpu_offload:", e)
|
63 |
|
64 |
+
# 5) VAE slicing reduces peak memory usage
|
65 |
pipe_local.enable_vae_slicing()
|
66 |
|
67 |
+
# Save to global
|
68 |
+
pipe_local.max_sequence_length = 256
|
69 |
pipe = pipe_local
|
70 |
+
print("Flux pipeline loaded successfully.")
|
71 |
|
72 |
###############################################################################
|
73 |
+
# Main inference function
|
74 |
###############################################################################
|
75 |
@spaces.GPU
|
76 |
def process_image(
|
77 |
+
image: Image.Image,
|
78 |
+
mask_image: Image.Image,
|
79 |
+
prompt="A person",
|
|
|
80 |
strength=0.75,
|
81 |
seed=0,
|
82 |
num_inference_steps=4,
|
83 |
progress=gr.Progress(track_tqdm=True)
|
84 |
):
|
85 |
"""
|
86 |
+
Loads the pipeline if needed, resizes the input image,
|
87 |
+
then runs Flux Img2Img with minimal VRAM usage strategies.
|
88 |
"""
|
89 |
+
progress(0, desc="Preparing model")
|
90 |
|
91 |
+
# 1) Ensure pipeline is loaded
|
92 |
+
load_flux_pipeline()
|
93 |
|
94 |
+
progress(20, desc="Resizing input image")
|
95 |
if image is None:
|
96 |
print("No input image provided.")
|
97 |
return None
|
98 |
|
99 |
+
# 2) Resize the input image to reduce VRAM usage
|
100 |
+
image = resize_image(image, max_dim=512)
|
|
|
|
|
|
|
101 |
|
102 |
+
# 3) Set up generator for reproducible results
|
103 |
generator = torch.Generator("cuda").manual_seed(seed)
|
104 |
+
|
105 |
+
# 4) Run the pipeline
|
106 |
+
progress(50, desc="Running Flux Inference")
|
107 |
+
print(f"Prompt: {prompt} | Strength: {strength} | Steps: {num_inference_steps}")
|
108 |
output = pipe(
|
109 |
prompt=prompt,
|
110 |
image=image,
|
111 |
generator=generator,
|
112 |
strength=strength,
|
113 |
+
guidance_scale=0, # matches your original code
|
114 |
num_inference_steps=num_inference_steps
|
115 |
)
|
|
|
116 |
|
117 |
+
progress(100, desc="Done")
|
118 |
return output.images[0]
|
119 |
|
120 |
###############################################################################
|
|
|
132 |
"""
|
133 |
|
134 |
with gr.Blocks(css=css) as demo:
|
135 |
+
gr.Markdown("## FLUX Img2Img — Memory-Optimized for T4\n"
|
136 |
+
"Using float16, CPU offload, xFormers, and image resizing to reduce VRAM usage.")
|
137 |
|
138 |
with gr.Row():
|
139 |
with gr.Column():
|
140 |
+
# The main input image
|
141 |
+
input_image = gr.Image(
|
142 |
label="Input Image (Img2Img)",
|
143 |
type="pil",
|
144 |
image_mode="RGB",
|
145 |
height=512
|
146 |
)
|
147 |
+
|
148 |
+
# Mask is not used in your code, but we keep it to match your function signature
|
149 |
+
mask_image = gr.Image(
|
150 |
label="Mask (unused)",
|
151 |
type="pil",
|
152 |
image_mode="RGB",
|
153 |
+
height=200
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
154 |
)
|
155 |
+
|
156 |
+
prompt = gr.Textbox(label="Prompt", value="A person")
|
157 |
+
strength_slider = gr.Slider(0.0, 1.0, value=0.75, step=0.05, label="Strength")
|
158 |
+
seed_box = gr.Number(value=0, label="Seed", precision=0)
|
159 |
+
steps_box = gr.Slider(1, 50, value=4, step=1, label="Inference Steps")
|
160 |
+
|
161 |
+
run_button = gr.Button("Generate")
|
162 |
|
163 |
with gr.Column():
|
164 |
+
result_image = gr.Image(
|
165 |
+
label="Output",
|
166 |
+
type="pil",
|
167 |
+
height=512
|
168 |
+
)
|
169 |
|
170 |
+
# Tie the button to our inference function
|
171 |
run_button.click(
|
172 |
fn=process_image,
|
173 |
+
inputs=[input_image, mask_image, prompt, strength_slider, seed_box, steps_box],
|
174 |
+
outputs=result_image
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
175 |
)
|
176 |
|
177 |
if __name__ == "__main__":
|