Commit
·
dc528d7
1
Parent(s):
2099bb6
support multimodel
Browse files
app.py
CHANGED
|
@@ -9,9 +9,12 @@ from pipeline.util import (
|
|
| 9 |
create_hdr_effect,
|
| 10 |
progressive_upscale,
|
| 11 |
select_scheduler,
|
|
|
|
| 12 |
)
|
| 13 |
|
| 14 |
device = "cuda"
|
|
|
|
|
|
|
| 15 |
|
| 16 |
# Initialize the models and pipeline
|
| 17 |
controlnet = ControlNetUnionModel.from_pretrained(
|
|
@@ -19,19 +22,28 @@ controlnet = ControlNetUnionModel.from_pretrained(
|
|
| 19 |
).to(device=device)
|
| 20 |
vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16).to(device=device)
|
| 21 |
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
model_id
|
| 26 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 27 |
|
| 28 |
-
|
| 29 |
-
pipe.enable_vae_tiling() # << Enable this if you have limited VRAM
|
| 30 |
-
pipe.enable_vae_slicing() # << Enable this if you have limited VRAM
|
| 31 |
|
| 32 |
# region functions
|
| 33 |
@spaces.GPU(duration=120)
|
| 34 |
def predict(
|
|
|
|
| 35 |
image,
|
| 36 |
prompt,
|
| 37 |
negative_prompt,
|
|
@@ -49,6 +61,9 @@ def predict(
|
|
| 49 |
):
|
| 50 |
global pipe
|
| 51 |
|
|
|
|
|
|
|
|
|
|
| 52 |
# Set selected scheduler
|
| 53 |
print(f"Using scheduler: {scheduler}...")
|
| 54 |
pipe.scheduler = select_scheduler(pipe, scheduler)
|
|
@@ -97,7 +112,6 @@ def predict(
|
|
| 97 |
|
| 98 |
return image
|
| 99 |
|
| 100 |
-
|
| 101 |
def clear_result():
|
| 102 |
return gr.update(value=None)
|
| 103 |
|
|
@@ -111,6 +125,38 @@ def set_maximum_resolution(max_tile_size, current_value):
|
|
| 111 |
def select_tile_weighting_method(tile_weighting_method):
|
| 112 |
return gr.update(visible=True if tile_weighting_method=="Gaussian" else False)
|
| 113 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 114 |
# endregion
|
| 115 |
|
| 116 |
css = """
|
|
@@ -207,6 +253,8 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
| 207 |
result = gr.Image(
|
| 208 |
label="Generated Image", show_label=True, format="png", interactive=False, scale=1, height=500, min_width=670
|
| 209 |
)
|
|
|
|
|
|
|
| 210 |
with gr.Row():
|
| 211 |
with gr.Accordion("Input Prompt", open=False):
|
| 212 |
with gr.Column():
|
|
@@ -231,6 +279,9 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
| 231 |
with gr.Sidebar(label="Parameters", open=True):
|
| 232 |
with gr.Row(elem_id="parameters_row"):
|
| 233 |
gr.Markdown("### General parameters")
|
|
|
|
|
|
|
|
|
|
| 234 |
tile_weighting_method = gr.Dropdown(
|
| 235 |
label="Tile Weighting Method", choices=["Cosine", "Gaussian"], value="Cosine"
|
| 236 |
)
|
|
@@ -260,7 +311,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
| 260 |
negative_prompt.value,
|
| 261 |
4096,
|
| 262 |
0.0,
|
| 263 |
-
|
| 264 |
0.35,
|
| 265 |
1.0,
|
| 266 |
0.3,
|
|
@@ -282,7 +333,21 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
| 282 |
4,
|
| 283 |
1024,
|
| 284 |
"Cosine"
|
| 285 |
-
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 286 |
[ "./examples/2.jpg",
|
| 287 |
prompt.value,
|
| 288 |
negative_prompt.value,
|
|
@@ -297,6 +362,20 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
| 297 |
1024,
|
| 298 |
"Cosine"
|
| 299 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 300 |
[ "./examples/3.jpg",
|
| 301 |
prompt.value,
|
| 302 |
negative_prompt.value,
|
|
@@ -311,6 +390,20 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
| 311 |
1280,
|
| 312 |
"Gaussian"
|
| 313 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 314 |
[ "./examples/4.jpg",
|
| 315 |
prompt.value,
|
| 316 |
negative_prompt.value,
|
|
@@ -325,6 +418,20 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
| 325 |
1024,
|
| 326 |
"Gaussian"
|
| 327 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 328 |
[ "./examples/5.jpg",
|
| 329 |
prompt.value,
|
| 330 |
negative_prompt.value,
|
|
@@ -338,7 +445,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
| 338 |
4,
|
| 339 |
1024,
|
| 340 |
"Cosine"
|
| 341 |
-
]
|
| 342 |
],
|
| 343 |
inputs=[
|
| 344 |
input_image,
|
|
@@ -355,7 +462,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
| 355 |
max_tile_size,
|
| 356 |
tile_weighting_method,
|
| 357 |
],
|
| 358 |
-
fn=
|
| 359 |
outputs=result,
|
| 360 |
cache_examples=False,
|
| 361 |
)
|
|
@@ -369,6 +476,7 @@ with gr.Blocks(css=css, theme=gr.themes.Ocean(), title="MoD ControlNet Tile Upsc
|
|
| 369 |
).then(
|
| 370 |
fn=predict,
|
| 371 |
inputs=[
|
|
|
|
| 372 |
input_image,
|
| 373 |
prompt,
|
| 374 |
negative_prompt,
|
|
|
|
| 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 |
# Initialize the models and pipeline
|
| 20 |
controlnet = ControlNetUnionModel.from_pretrained(
|
|
|
|
| 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 |
+
pipe = None
|
| 30 |
+
torch_gc()
|
| 31 |
+
|
| 32 |
+
pipe = StableDiffusionXLControlNetTileSRPipeline.from_pretrained(
|
| 33 |
+
model_id, controlnet=controlnet, vae=vae, torch_dtype=torch.float16, use_safetensors=True, variant="fp16"
|
| 34 |
+
).to(device)
|
| 35 |
+
|
| 36 |
+
#pipe.enable_model_cpu_offload() # << Enable this if you have limited VRAM
|
| 37 |
+
pipe.enable_vae_tiling() # << Enable this if you have limited VRAM
|
| 38 |
+
pipe.enable_vae_slicing() # << Enable this if you have limited VRAM
|
| 39 |
+
last_loaded_model = model_id
|
| 40 |
|
| 41 |
+
load_model("SG161222/RealVisXL_V5.0_Lightning")
|
|
|
|
|
|
|
| 42 |
|
| 43 |
# region functions
|
| 44 |
@spaces.GPU(duration=120)
|
| 45 |
def predict(
|
| 46 |
+
model_id,
|
| 47 |
image,
|
| 48 |
prompt,
|
| 49 |
negative_prompt,
|
|
|
|
| 61 |
):
|
| 62 |
global pipe
|
| 63 |
|
| 64 |
+
# Load model if changed
|
| 65 |
+
load_model(model_id)
|
| 66 |
+
|
| 67 |
# Set selected scheduler
|
| 68 |
print(f"Using scheduler: {scheduler}...")
|
| 69 |
pipe.scheduler = select_scheduler(pipe, scheduler)
|
|
|
|
| 112 |
|
| 113 |
return image
|
| 114 |
|
|
|
|
| 115 |
def clear_result():
|
| 116 |
return gr.update(value=None)
|
| 117 |
|
|
|
|
| 125 |
def select_tile_weighting_method(tile_weighting_method):
|
| 126 |
return gr.update(visible=True if tile_weighting_method=="Gaussian" else False)
|
| 127 |
|
| 128 |
+
@spaces.GPU(duration=120)
|
| 129 |
+
def run_for_examples(image,
|
| 130 |
+
prompt,
|
| 131 |
+
negative_prompt,
|
| 132 |
+
resolution,
|
| 133 |
+
hdr,
|
| 134 |
+
num_inference_steps,
|
| 135 |
+
denoising_strenght,
|
| 136 |
+
controlnet_strength,
|
| 137 |
+
tile_gaussian_sigma,
|
| 138 |
+
scheduler,
|
| 139 |
+
guidance_scale,
|
| 140 |
+
max_tile_size,
|
| 141 |
+
tile_weighting_method):
|
| 142 |
+
|
| 143 |
+
predict(
|
| 144 |
+
model.value,
|
| 145 |
+
image,
|
| 146 |
+
prompt,
|
| 147 |
+
negative_prompt,
|
| 148 |
+
resolution,
|
| 149 |
+
hdr,
|
| 150 |
+
num_inference_steps,
|
| 151 |
+
denoising_strenght,
|
| 152 |
+
controlnet_strength,
|
| 153 |
+
tile_gaussian_sigma,
|
| 154 |
+
scheduler,
|
| 155 |
+
guidance_scale,
|
| 156 |
+
max_tile_size,
|
| 157 |
+
tile_weighting_method)
|
| 158 |
+
|
| 159 |
+
|
| 160 |
# endregion
|
| 161 |
|
| 162 |
css = """
|
|
|
|
| 253 |
result = gr.Image(
|
| 254 |
label="Generated Image", show_label=True, format="png", interactive=False, scale=1, height=500, min_width=670
|
| 255 |
)
|
| 256 |
+
with gr.Row():
|
| 257 |
+
gr.HTML("<div style='color: red;'>If you are not a Pro account, run the LCM sampler examples on the RealVisXL_V5.0_Lightning model. For best results use the UniPC sampler and RealVisXL_V5.0 model examples.</div>")
|
| 258 |
with gr.Row():
|
| 259 |
with gr.Accordion("Input Prompt", open=False):
|
| 260 |
with gr.Column():
|
|
|
|
| 279 |
with gr.Sidebar(label="Parameters", open=True):
|
| 280 |
with gr.Row(elem_id="parameters_row"):
|
| 281 |
gr.Markdown("### General parameters")
|
| 282 |
+
model = gr.Dropdown(
|
| 283 |
+
label="Model", choices=["SG161222/RealVisXL_V5.0_Lightning", "SG161222/RealVisXL_V5.0"], value="SG161222/RealVisXL_V5.0_Lightning"
|
| 284 |
+
)
|
| 285 |
tile_weighting_method = gr.Dropdown(
|
| 286 |
label="Tile Weighting Method", choices=["Cosine", "Gaussian"], value="Cosine"
|
| 287 |
)
|
|
|
|
| 311 |
negative_prompt.value,
|
| 312 |
4096,
|
| 313 |
0.0,
|
| 314 |
+
25,
|
| 315 |
0.35,
|
| 316 |
1.0,
|
| 317 |
0.3,
|
|
|
|
| 333 |
4,
|
| 334 |
1024,
|
| 335 |
"Cosine"
|
| 336 |
+
],
|
| 337 |
+
[ "./examples/2.jpg",
|
| 338 |
+
prompt.value,
|
| 339 |
+
negative_prompt.value,
|
| 340 |
+
4096,
|
| 341 |
+
0.5,
|
| 342 |
+
25,
|
| 343 |
+
0.35,
|
| 344 |
+
1.0,
|
| 345 |
+
0.3,
|
| 346 |
+
"LCM",
|
| 347 |
+
4,
|
| 348 |
+
1024,
|
| 349 |
+
"Cosine"
|
| 350 |
+
],
|
| 351 |
[ "./examples/2.jpg",
|
| 352 |
prompt.value,
|
| 353 |
negative_prompt.value,
|
|
|
|
| 362 |
1024,
|
| 363 |
"Cosine"
|
| 364 |
],
|
| 365 |
+
[ "./examples/3.jpg",
|
| 366 |
+
prompt.value,
|
| 367 |
+
negative_prompt.value,
|
| 368 |
+
5120,
|
| 369 |
+
0.5,
|
| 370 |
+
25,
|
| 371 |
+
0.35,
|
| 372 |
+
1.0,
|
| 373 |
+
0.3,
|
| 374 |
+
"LCM",
|
| 375 |
+
4,
|
| 376 |
+
1280,
|
| 377 |
+
"Gaussian"
|
| 378 |
+
],
|
| 379 |
[ "./examples/3.jpg",
|
| 380 |
prompt.value,
|
| 381 |
negative_prompt.value,
|
|
|
|
| 390 |
1280,
|
| 391 |
"Gaussian"
|
| 392 |
],
|
| 393 |
+
[ "./examples/4.jpg",
|
| 394 |
+
prompt.value,
|
| 395 |
+
negative_prompt.value,
|
| 396 |
+
8192,
|
| 397 |
+
0.1,
|
| 398 |
+
25,
|
| 399 |
+
0.35,
|
| 400 |
+
1.0,
|
| 401 |
+
0.3,
|
| 402 |
+
"LCM",
|
| 403 |
+
4,
|
| 404 |
+
1024,
|
| 405 |
+
"Gaussian"
|
| 406 |
+
],
|
| 407 |
[ "./examples/4.jpg",
|
| 408 |
prompt.value,
|
| 409 |
negative_prompt.value,
|
|
|
|
| 418 |
1024,
|
| 419 |
"Gaussian"
|
| 420 |
],
|
| 421 |
+
[ "./examples/5.jpg",
|
| 422 |
+
prompt.value,
|
| 423 |
+
negative_prompt.value,
|
| 424 |
+
8192,
|
| 425 |
+
0.3,
|
| 426 |
+
25,
|
| 427 |
+
0.35,
|
| 428 |
+
1.0,
|
| 429 |
+
0.3,
|
| 430 |
+
"LCM",
|
| 431 |
+
4,
|
| 432 |
+
1024,
|
| 433 |
+
"Cosine"
|
| 434 |
+
],
|
| 435 |
[ "./examples/5.jpg",
|
| 436 |
prompt.value,
|
| 437 |
negative_prompt.value,
|
|
|
|
| 445 |
4,
|
| 446 |
1024,
|
| 447 |
"Cosine"
|
| 448 |
+
]
|
| 449 |
],
|
| 450 |
inputs=[
|
| 451 |
input_image,
|
|
|
|
| 462 |
max_tile_size,
|
| 463 |
tile_weighting_method,
|
| 464 |
],
|
| 465 |
+
fn=run_for_examples,
|
| 466 |
outputs=result,
|
| 467 |
cache_examples=False,
|
| 468 |
)
|
|
|
|
| 476 |
).then(
|
| 477 |
fn=predict,
|
| 478 |
inputs=[
|
| 479 |
+
model,
|
| 480 |
input_image,
|
| 481 |
prompt,
|
| 482 |
negative_prompt,
|