Spaces:
Running
on
Zero
Running
on
Zero
Commit
·
6138235
1
Parent(s):
14638d9
switch model
Browse files- app.py +44 -55
- pipeline/util.py +29 -1
app.py
CHANGED
@@ -1,49 +1,59 @@
|
|
1 |
import torch
|
2 |
import spaces
|
3 |
-
from diffusers import ControlNetUnionModel, AutoencoderKL
|
4 |
import gradio as gr
|
5 |
|
6 |
from pipeline.mod_controlnet_tile_sr_sdxl import StableDiffusionXLControlNetTileSRPipeline, calculate_overlap
|
7 |
from pipeline.util import (
|
8 |
SAMPLERS,
|
9 |
create_hdr_effect,
|
|
|
10 |
progressive_upscale,
|
|
|
11 |
select_scheduler,
|
12 |
-
torch_gc,
|
13 |
)
|
14 |
|
15 |
device = "cuda"
|
16 |
pipe = None
|
17 |
last_loaded_model = None
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
"brad-twinkl/controlnet-union-sdxl-1.0-promax", torch_dtype=torch.float16
|
22 |
-
).to(device=device)
|
23 |
-
vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16).to(device=device)
|
24 |
|
25 |
def load_model(model_id):
|
26 |
global pipe, last_loaded_model
|
27 |
-
|
28 |
if model_id != last_loaded_model:
|
29 |
-
|
30 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
31 |
pipe = StableDiffusionXLControlNetTileSRPipeline.from_pretrained(
|
32 |
-
model_id, controlnet=controlnet, vae=vae, torch_dtype=torch.float16, use_safetensors=True, variant="fp16"
|
33 |
-
)
|
34 |
-
|
35 |
-
#pipe.enable_model_cpu_offload() # << Enable this if you have limited VRAM
|
36 |
pipe.enable_vae_tiling() # << Enable this if you have limited VRAM
|
37 |
pipe.enable_vae_slicing() # << Enable this if you have limited VRAM
|
|
|
|
|
|
|
|
|
|
|
38 |
last_loaded_model = model_id
|
39 |
|
40 |
-
load_model("
|
41 |
|
42 |
# region functions
|
43 |
@spaces.GPU(duration=120)
|
44 |
def predict(
|
45 |
-
model_id,
|
46 |
image,
|
|
|
47 |
prompt,
|
48 |
negative_prompt,
|
49 |
resolution,
|
@@ -124,38 +134,6 @@ def set_maximum_resolution(max_tile_size, current_value):
|
|
124 |
def select_tile_weighting_method(tile_weighting_method):
|
125 |
return gr.update(visible=True if tile_weighting_method=="Gaussian" else False)
|
126 |
|
127 |
-
@spaces.GPU(duration=120)
|
128 |
-
def run_for_examples(image,
|
129 |
-
prompt,
|
130 |
-
negative_prompt,
|
131 |
-
resolution,
|
132 |
-
hdr,
|
133 |
-
num_inference_steps,
|
134 |
-
denoising_strenght,
|
135 |
-
controlnet_strength,
|
136 |
-
tile_gaussian_sigma,
|
137 |
-
scheduler,
|
138 |
-
guidance_scale,
|
139 |
-
max_tile_size,
|
140 |
-
tile_weighting_method):
|
141 |
-
|
142 |
-
predict(
|
143 |
-
model.value,
|
144 |
-
image,
|
145 |
-
prompt,
|
146 |
-
negative_prompt,
|
147 |
-
resolution,
|
148 |
-
hdr,
|
149 |
-
num_inference_steps,
|
150 |
-
denoising_strenght,
|
151 |
-
controlnet_strength,
|
152 |
-
tile_gaussian_sigma,
|
153 |
-
scheduler,
|
154 |
-
guidance_scale,
|
155 |
-
max_tile_size,
|
156 |
-
tile_weighting_method)
|
157 |
-
|
158 |
-
|
159 |
# endregion
|
160 |
|
161 |
css = """
|
@@ -174,7 +152,7 @@ body {
|
|
174 |
text-shadow: 1px 1px 2px rgba(0, 0, 0, 0.2);
|
175 |
}
|
176 |
.fillable {
|
177 |
-
width:
|
178 |
max-width: unset !important;
|
179 |
}
|
180 |
#examples_container {
|
@@ -279,7 +257,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
279 |
with gr.Row(elem_id="parameters_row"):
|
280 |
gr.Markdown("### General parameters")
|
281 |
model = gr.Dropdown(
|
282 |
-
label="Model", choices=
|
283 |
)
|
284 |
tile_weighting_method = gr.Dropdown(
|
285 |
label="Tile Weighting Method", choices=["Cosine", "Gaussian"], value="Cosine"
|
@@ -303,9 +281,10 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
303 |
with gr.Accordion(label="Example Images", open=True):
|
304 |
with gr.Row(elem_id="examples_row"):
|
305 |
with gr.Column(scale=12, elem_id="examples_container"):
|
306 |
-
gr.Examples(
|
307 |
examples=[
|
308 |
[ "./examples/1.jpg",
|
|
|
309 |
prompt.value,
|
310 |
negative_prompt.value,
|
311 |
4096,
|
@@ -320,6 +299,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
320 |
"Cosine"
|
321 |
],
|
322 |
[ "./examples/1.jpg",
|
|
|
323 |
prompt.value,
|
324 |
negative_prompt.value,
|
325 |
4096,
|
@@ -334,6 +314,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
334 |
"Cosine"
|
335 |
],
|
336 |
[ "./examples/2.jpg",
|
|
|
337 |
prompt.value,
|
338 |
negative_prompt.value,
|
339 |
4096,
|
@@ -348,6 +329,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
348 |
"Cosine"
|
349 |
],
|
350 |
[ "./examples/2.jpg",
|
|
|
351 |
prompt.value,
|
352 |
negative_prompt.value,
|
353 |
4096,
|
@@ -362,6 +344,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
362 |
"Cosine"
|
363 |
],
|
364 |
[ "./examples/3.jpg",
|
|
|
365 |
prompt.value,
|
366 |
negative_prompt.value,
|
367 |
5120,
|
@@ -376,6 +359,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
376 |
"Gaussian"
|
377 |
],
|
378 |
[ "./examples/3.jpg",
|
|
|
379 |
prompt.value,
|
380 |
negative_prompt.value,
|
381 |
5120,
|
@@ -390,6 +374,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
390 |
"Gaussian"
|
391 |
],
|
392 |
[ "./examples/4.jpg",
|
|
|
393 |
prompt.value,
|
394 |
negative_prompt.value,
|
395 |
8192,
|
@@ -404,6 +389,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
404 |
"Gaussian"
|
405 |
],
|
406 |
[ "./examples/4.jpg",
|
|
|
407 |
prompt.value,
|
408 |
negative_prompt.value,
|
409 |
8192,
|
@@ -418,6 +404,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
418 |
"Gaussian"
|
419 |
],
|
420 |
[ "./examples/5.jpg",
|
|
|
421 |
prompt.value,
|
422 |
negative_prompt.value,
|
423 |
8192,
|
@@ -432,6 +419,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
432 |
"Cosine"
|
433 |
],
|
434 |
[ "./examples/5.jpg",
|
|
|
435 |
prompt.value,
|
436 |
negative_prompt.value,
|
437 |
8192,
|
@@ -448,6 +436,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
448 |
],
|
449 |
inputs=[
|
450 |
input_image,
|
|
|
451 |
prompt,
|
452 |
negative_prompt,
|
453 |
resolution,
|
@@ -461,13 +450,13 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
461 |
max_tile_size,
|
462 |
tile_weighting_method,
|
463 |
],
|
464 |
-
fn=
|
465 |
outputs=result,
|
466 |
cache_examples=False,
|
467 |
)
|
468 |
|
469 |
max_tile_size.select(fn=set_maximum_resolution, inputs=[max_tile_size, resolution], outputs=resolution)
|
470 |
-
tile_weighting_method.
|
471 |
generate_button.click(
|
472 |
fn=clear_result,
|
473 |
inputs=None,
|
@@ -475,8 +464,8 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
475 |
).then(
|
476 |
fn=predict,
|
477 |
inputs=[
|
478 |
-
model,
|
479 |
input_image,
|
|
|
480 |
prompt,
|
481 |
negative_prompt,
|
482 |
resolution,
|
|
|
1 |
import torch
|
2 |
import spaces
|
3 |
+
from diffusers import ControlNetUnionModel, AutoencoderKL, UNet2DConditionModel
|
4 |
import gradio as gr
|
5 |
|
6 |
from pipeline.mod_controlnet_tile_sr_sdxl import StableDiffusionXLControlNetTileSRPipeline, calculate_overlap
|
7 |
from pipeline.util import (
|
8 |
SAMPLERS,
|
9 |
create_hdr_effect,
|
10 |
+
optionally_disable_offloading,
|
11 |
progressive_upscale,
|
12 |
+
quantize_8bit,
|
13 |
select_scheduler,
|
14 |
+
torch_gc,
|
15 |
)
|
16 |
|
17 |
device = "cuda"
|
18 |
pipe = None
|
19 |
last_loaded_model = None
|
20 |
+
MODELS = {"RealVisXL 5 Lightning": "SG161222/RealVisXL_V5.0_Lightning",
|
21 |
+
"RealVisXL 5": "SG161222/RealVisXL_V5.0"
|
22 |
+
}
|
|
|
|
|
|
|
23 |
|
24 |
def load_model(model_id):
|
25 |
global pipe, last_loaded_model
|
26 |
+
|
27 |
if model_id != last_loaded_model:
|
28 |
+
|
29 |
+
# Initialize the models and pipeline
|
30 |
+
controlnet = ControlNetUnionModel.from_pretrained(
|
31 |
+
"brad-twinkl/controlnet-union-sdxl-1.0-promax", torch_dtype=torch.float16
|
32 |
+
)
|
33 |
+
vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
|
34 |
+
if pipe is not None:
|
35 |
+
optionally_disable_offloading(pipe)
|
36 |
+
torch_gc()
|
37 |
pipe = StableDiffusionXLControlNetTileSRPipeline.from_pretrained(
|
38 |
+
MODELS[model_id], controlnet=controlnet, vae=vae, torch_dtype=torch.float16, use_safetensors=True, variant="fp16"
|
39 |
+
)
|
40 |
+
pipe.enable_model_cpu_offload() # << Enable this if you have limited VRAM
|
|
|
41 |
pipe.enable_vae_tiling() # << Enable this if you have limited VRAM
|
42 |
pipe.enable_vae_slicing() # << Enable this if you have limited VRAM
|
43 |
+
|
44 |
+
unet = UNet2DConditionModel.from_pretrained(MODELS[model_id], subfolder="unet", variant="fp16", use_safetensors=True)
|
45 |
+
quantize_8bit(unet) # << Enable this if you have limited VRAM
|
46 |
+
pipe.unet = unet
|
47 |
+
|
48 |
last_loaded_model = model_id
|
49 |
|
50 |
+
load_model("RealVisXL 5 Lightning")
|
51 |
|
52 |
# region functions
|
53 |
@spaces.GPU(duration=120)
|
54 |
def predict(
|
|
|
55 |
image,
|
56 |
+
model_id,
|
57 |
prompt,
|
58 |
negative_prompt,
|
59 |
resolution,
|
|
|
134 |
def select_tile_weighting_method(tile_weighting_method):
|
135 |
return gr.update(visible=True if tile_weighting_method=="Gaussian" else False)
|
136 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
137 |
# endregion
|
138 |
|
139 |
css = """
|
|
|
152 |
text-shadow: 1px 1px 2px rgba(0, 0, 0, 0.2);
|
153 |
}
|
154 |
.fillable {
|
155 |
+
width: 100% !important;
|
156 |
max-width: unset !important;
|
157 |
}
|
158 |
#examples_container {
|
|
|
257 |
with gr.Row(elem_id="parameters_row"):
|
258 |
gr.Markdown("### General parameters")
|
259 |
model = gr.Dropdown(
|
260 |
+
label="Model", choices=MODELS.keys(), value=list(MODELS.keys())[0]
|
261 |
)
|
262 |
tile_weighting_method = gr.Dropdown(
|
263 |
label="Tile Weighting Method", choices=["Cosine", "Gaussian"], value="Cosine"
|
|
|
281 |
with gr.Accordion(label="Example Images", open=True):
|
282 |
with gr.Row(elem_id="examples_row"):
|
283 |
with gr.Column(scale=12, elem_id="examples_container"):
|
284 |
+
eg = gr.Examples(
|
285 |
examples=[
|
286 |
[ "./examples/1.jpg",
|
287 |
+
"RealVisXL 5 Lightning",
|
288 |
prompt.value,
|
289 |
negative_prompt.value,
|
290 |
4096,
|
|
|
299 |
"Cosine"
|
300 |
],
|
301 |
[ "./examples/1.jpg",
|
302 |
+
"RealVisXL 5",
|
303 |
prompt.value,
|
304 |
negative_prompt.value,
|
305 |
4096,
|
|
|
314 |
"Cosine"
|
315 |
],
|
316 |
[ "./examples/2.jpg",
|
317 |
+
"RealVisXL 5 Lightning",
|
318 |
prompt.value,
|
319 |
negative_prompt.value,
|
320 |
4096,
|
|
|
329 |
"Cosine"
|
330 |
],
|
331 |
[ "./examples/2.jpg",
|
332 |
+
"RealVisXL 5",
|
333 |
prompt.value,
|
334 |
negative_prompt.value,
|
335 |
4096,
|
|
|
344 |
"Cosine"
|
345 |
],
|
346 |
[ "./examples/3.jpg",
|
347 |
+
"RealVisXL 5 Lightning",
|
348 |
prompt.value,
|
349 |
negative_prompt.value,
|
350 |
5120,
|
|
|
359 |
"Gaussian"
|
360 |
],
|
361 |
[ "./examples/3.jpg",
|
362 |
+
"RealVisXL 5",
|
363 |
prompt.value,
|
364 |
negative_prompt.value,
|
365 |
5120,
|
|
|
374 |
"Gaussian"
|
375 |
],
|
376 |
[ "./examples/4.jpg",
|
377 |
+
"RealVisXL 5 Lightning",
|
378 |
prompt.value,
|
379 |
negative_prompt.value,
|
380 |
8192,
|
|
|
389 |
"Gaussian"
|
390 |
],
|
391 |
[ "./examples/4.jpg",
|
392 |
+
"RealVisXL 5",
|
393 |
prompt.value,
|
394 |
negative_prompt.value,
|
395 |
8192,
|
|
|
404 |
"Gaussian"
|
405 |
],
|
406 |
[ "./examples/5.jpg",
|
407 |
+
"RealVisXL 5 Lightning",
|
408 |
prompt.value,
|
409 |
negative_prompt.value,
|
410 |
8192,
|
|
|
419 |
"Cosine"
|
420 |
],
|
421 |
[ "./examples/5.jpg",
|
422 |
+
"RealVisXL 5",
|
423 |
prompt.value,
|
424 |
negative_prompt.value,
|
425 |
8192,
|
|
|
436 |
],
|
437 |
inputs=[
|
438 |
input_image,
|
439 |
+
model,
|
440 |
prompt,
|
441 |
negative_prompt,
|
442 |
resolution,
|
|
|
450 |
max_tile_size,
|
451 |
tile_weighting_method,
|
452 |
],
|
453 |
+
fn=predict,
|
454 |
outputs=result,
|
455 |
cache_examples=False,
|
456 |
)
|
457 |
|
458 |
max_tile_size.select(fn=set_maximum_resolution, inputs=[max_tile_size, resolution], outputs=resolution)
|
459 |
+
tile_weighting_method.change(fn=select_tile_weighting_method, inputs=tile_weighting_method, outputs=tile_gaussian_sigma)
|
460 |
generate_button.click(
|
461 |
fn=clear_result,
|
462 |
inputs=None,
|
|
|
464 |
).then(
|
465 |
fn=predict,
|
466 |
inputs=[
|
|
|
467 |
input_image,
|
468 |
+
model,
|
469 |
prompt,
|
470 |
negative_prompt,
|
471 |
resolution,
|
pipeline/util.py
CHANGED
@@ -16,6 +16,8 @@
|
|
16 |
import gc
|
17 |
import cv2
|
18 |
import numpy as np
|
|
|
|
|
19 |
import torch
|
20 |
from PIL import Image
|
21 |
|
@@ -96,6 +98,32 @@ def select_scheduler(pipe, selected_sampler):
|
|
96 |
|
97 |
return scheduler.from_config(config, **add_kwargs)
|
98 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
99 |
|
100 |
# This function was copied and adapted from https://huggingface.co/spaces/gokaygokay/TileUpscalerV2, licensed under Apache 2.0.
|
101 |
def progressive_upscale(input_image, target_resolution, steps=3):
|
@@ -185,7 +213,7 @@ def torch_gc():
|
|
185 |
if torch.cuda.is_available():
|
186 |
with torch.cuda.device("cuda"):
|
187 |
torch.cuda.empty_cache()
|
188 |
-
torch.cuda.ipc_collect()
|
189 |
|
190 |
gc.collect()
|
191 |
|
|
|
16 |
import gc
|
17 |
import cv2
|
18 |
import numpy as np
|
19 |
+
from torch import nn
|
20 |
+
from accelerate.hooks import AlignDevicesHook, CpuOffload, remove_hook_from_module
|
21 |
import torch
|
22 |
from PIL import Image
|
23 |
|
|
|
98 |
|
99 |
return scheduler.from_config(config, **add_kwargs)
|
100 |
|
101 |
+
def optionally_disable_offloading(_pipeline):
|
102 |
+
"""
|
103 |
+
Optionally removes offloading in case the pipeline has been already sequentially offloaded to CPU.
|
104 |
+
|
105 |
+
Args:
|
106 |
+
_pipeline (`DiffusionPipeline`):
|
107 |
+
The pipeline to disable offloading for.
|
108 |
+
|
109 |
+
Returns:
|
110 |
+
tuple:
|
111 |
+
A tuple indicating if `is_model_cpu_offload` or `is_sequential_cpu_offload` is True.
|
112 |
+
"""
|
113 |
+
is_model_cpu_offload = False
|
114 |
+
is_sequential_cpu_offload = False
|
115 |
+
if _pipeline is not None:
|
116 |
+
for _, component in _pipeline.components.items():
|
117 |
+
if isinstance(component, nn.Module) and hasattr(component, "_hf_hook"):
|
118 |
+
if not is_model_cpu_offload:
|
119 |
+
is_model_cpu_offload = isinstance(component._hf_hook, CpuOffload)
|
120 |
+
if not is_sequential_cpu_offload:
|
121 |
+
is_sequential_cpu_offload = isinstance(component._hf_hook, AlignDevicesHook)
|
122 |
+
|
123 |
+
|
124 |
+
remove_hook_from_module(component, recurse=True)
|
125 |
+
|
126 |
+
return (is_model_cpu_offload, is_sequential_cpu_offload)
|
127 |
|
128 |
# This function was copied and adapted from https://huggingface.co/spaces/gokaygokay/TileUpscalerV2, licensed under Apache 2.0.
|
129 |
def progressive_upscale(input_image, target_resolution, steps=3):
|
|
|
213 |
if torch.cuda.is_available():
|
214 |
with torch.cuda.device("cuda"):
|
215 |
torch.cuda.empty_cache()
|
216 |
+
#torch.cuda.ipc_collect()
|
217 |
|
218 |
gc.collect()
|
219 |
|