final model
Browse files
app.py
CHANGED
|
@@ -243,258 +243,6 @@ gr_interface = gr.Interface(
|
|
| 243 |
|
| 244 |
|
| 245 |
if __name__ == "__main__":
|
| 246 |
-
|
| 247 |
-
|
| 248 |
-
gr_interface.queue(15).launch()
|
| 249 |
-
|
| 250 |
-
# theme = gr.themes.Soft(
|
| 251 |
-
# primary_hue="blue",
|
| 252 |
-
# secondary_hue="slate",
|
| 253 |
-
# neutral_hue="slate",
|
| 254 |
-
# text_size="sm",
|
| 255 |
-
# spacing_size="sm",
|
| 256 |
-
# ).set(
|
| 257 |
-
# background_fill_primary="*neutral_50",
|
| 258 |
-
# background_fill_secondary="*background_fill_primary",
|
| 259 |
-
# block_background_fill="*background_fill_primary",
|
| 260 |
-
# block_border_width="0px",
|
| 261 |
-
# panel_background_fill="*neutral_50",
|
| 262 |
-
# panel_border_width="0px",
|
| 263 |
-
# input_background_fill="*neutral_100",
|
| 264 |
-
# input_border_color="*neutral_200",
|
| 265 |
-
# button_primary_background_fill="*primary_300",
|
| 266 |
-
# button_primary_background_fill_hover="*primary_400",
|
| 267 |
-
# button_secondary_background_fill="*neutral_200",
|
| 268 |
-
# button_secondary_background_fill_hover="*neutral_300",
|
| 269 |
-
# )
|
| 270 |
-
# custom_css = """
|
| 271 |
-
# #main-headertitle {
|
| 272 |
-
# text-align: center;
|
| 273 |
-
# margin-top: 15px;
|
| 274 |
-
# margin-bottom: 10px;
|
| 275 |
-
# color: var(--neutral-600);
|
| 276 |
-
# font-weight: 600;
|
| 277 |
-
# }
|
| 278 |
-
# #main-header {
|
| 279 |
-
# text-align: center;
|
| 280 |
-
# margin-top: 5px;
|
| 281 |
-
# margin-bottom: 10px;
|
| 282 |
-
# color: var(--neutral-600);
|
| 283 |
-
# font-weight: 600;
|
| 284 |
-
# }
|
| 285 |
-
# #model-settings-header, #generation-settings-header {
|
| 286 |
-
# color: var(--neutral-600);
|
| 287 |
-
# margin-top: 8px;
|
| 288 |
-
# margin-bottom: 8px;
|
| 289 |
-
# font-weight: 500;
|
| 290 |
-
# font-size: 1.1em;
|
| 291 |
-
# }
|
| 292 |
-
# .setting-section {
|
| 293 |
-
# padding: 10px 12px;
|
| 294 |
-
# border-radius: 6px;
|
| 295 |
-
# background-color: var(--neutral-50);
|
| 296 |
-
# margin-bottom: 10px;
|
| 297 |
-
# border: 1px solid var(--neutral-100);
|
| 298 |
-
# }
|
| 299 |
-
# hr {
|
| 300 |
-
# border: none;
|
| 301 |
-
# height: 1px;
|
| 302 |
-
# background-color: var(--neutral-200);
|
| 303 |
-
# margin: 8px 0;
|
| 304 |
-
# }
|
| 305 |
-
# #generate-btn {
|
| 306 |
-
# width: 100%;
|
| 307 |
-
# max-width: 250px;
|
| 308 |
-
# margin: 10px auto;
|
| 309 |
-
# display: block;
|
| 310 |
-
# padding: 10px 15px;
|
| 311 |
-
# font-size: 16px;
|
| 312 |
-
# border-radius: 5px;
|
| 313 |
-
# }
|
| 314 |
-
# #status-box {
|
| 315 |
-
# min-height: 50px;
|
| 316 |
-
# display: flex;
|
| 317 |
-
# align-items: center;
|
| 318 |
-
# justify-content: center;
|
| 319 |
-
# padding: 8px;
|
| 320 |
-
# border-radius: 5px;
|
| 321 |
-
# border: 1px solid var(--neutral-200);
|
| 322 |
-
# color: var(--neutral-700);
|
| 323 |
-
# }
|
| 324 |
-
# #project-badges {
|
| 325 |
-
# text-align: center;
|
| 326 |
-
# margin-top: 30px;
|
| 327 |
-
# margin-bottom: 20px;
|
| 328 |
-
# }
|
| 329 |
-
# #project-badges #badge-container {
|
| 330 |
-
# display: flex;
|
| 331 |
-
# gap: 10px;
|
| 332 |
-
# align-items: center;
|
| 333 |
-
# justify-content: center;
|
| 334 |
-
# flex-wrap: wrap;
|
| 335 |
-
# }
|
| 336 |
-
# #project-badges img {
|
| 337 |
-
# border-radius: 5px;
|
| 338 |
-
# box-shadow: 0 1px 3px rgba(0, 0, 0, 0.1);
|
| 339 |
-
# height: 20px;
|
| 340 |
-
# transition: transform 0.1s ease, box-shadow 0.1s ease;
|
| 341 |
-
# }
|
| 342 |
-
# #project-badges a:hover img {
|
| 343 |
-
# transform: translateY(-2px);
|
| 344 |
-
# box-shadow: 0 4px 8px rgba(0, 0, 0, 0.15);
|
| 345 |
-
# }
|
| 346 |
-
# #audio-output {
|
| 347 |
-
# height: 200px;
|
| 348 |
-
# border-radius: 5px;
|
| 349 |
-
# border: 1px solid var(--neutral-200);
|
| 350 |
-
# }
|
| 351 |
-
# .gradio-dropdown label, .gradio-checkbox label, .gradio-number label, .gradio-textbox label {
|
| 352 |
-
# font-weight: 500;
|
| 353 |
-
# color: var(--neutral-700);
|
| 354 |
-
# font-size: 0.9em;
|
| 355 |
-
# }
|
| 356 |
-
# .gradio-row {
|
| 357 |
-
# gap: 8px;
|
| 358 |
-
# }
|
| 359 |
-
# .gradio-block {
|
| 360 |
-
# margin-bottom: 8px;
|
| 361 |
-
# }
|
| 362 |
-
# .setting-section .gradio-block {
|
| 363 |
-
# margin-bottom: 6px;
|
| 364 |
-
# }
|
| 365 |
-
# ::-webkit-scrollbar {
|
| 366 |
-
# width: 8px;
|
| 367 |
-
# height: 8px;
|
| 368 |
-
# }
|
| 369 |
-
# ::-webkit-scrollbar-track {
|
| 370 |
-
# background: var(--neutral-100);
|
| 371 |
-
# border-radius: 4px;
|
| 372 |
-
# }
|
| 373 |
-
# ::-webkit-scrollbar-thumb {
|
| 374 |
-
# background: var(--neutral-300);
|
| 375 |
-
# border-radius: 4px;
|
| 376 |
-
# }
|
| 377 |
-
# ::-webkit-scrollbar-thumb:hover {
|
| 378 |
-
# background: var(--neutral-400);
|
| 379 |
-
# }
|
| 380 |
-
# * {
|
| 381 |
-
# scrollbar-width: thin;
|
| 382 |
-
# scrollbar-color: var(--neutral-300) var(--neutral-100);
|
| 383 |
-
# }
|
| 384 |
-
# """
|
| 385 |
-
# with gr.Blocks(title="MeanAudio Generator", theme=theme, css=custom_css) as demo:
|
| 386 |
-
# gr.Markdown("# MeanAudio: Fast and Faithful Text-to-Audio Generation with Mean Flows", elem_id="main-header")
|
| 387 |
-
# badge_html = '''
|
| 388 |
-
# <div id="project-badges"> <!-- 使用 ID
|
| 389 |
-
# 以便应用 CSS -->
|
| 390 |
-
# <div id="badge-container"> <!-- 添加这个容器 div 并使用 ID -->
|
| 391 |
-
# <a href="https://huggingface.co/junxiliu/MeanAudio">
|
| 392 |
-
# <img src="https://img.shields.io/badge/Model-HuggingFace-violet?logo=huggingface" alt="Hugging Face Model">
|
| 393 |
-
# </a>
|
| 394 |
-
# <a href="https://huggingface.co/spaces/chenxie95/MeanAudio">
|
| 395 |
-
# <img src="https://img.shields.io/badge/Space-HuggingFace-8A2BE2?logo=huggingface" alt="Hugging Face Space">
|
| 396 |
-
# </a>
|
| 397 |
-
# <a href="https://meanaudio.github.io/">
|
| 398 |
-
# <img src="https://img.shields.io/badge/Project-Page-brightred?style=flat" alt="Project Page">
|
| 399 |
-
# </a>
|
| 400 |
-
# <a href="https://github.com/xiquan-li/MeanAudio">
|
| 401 |
-
# <img src="https://img.shields.io/badge/Code-GitHub-black?logo=github" alt="GitHub">
|
| 402 |
-
# </a>
|
| 403 |
-
# </div>
|
| 404 |
-
# </div>
|
| 405 |
-
# '''
|
| 406 |
-
# gr.HTML(badge_html)
|
| 407 |
-
# with gr.Column(elem_classes="setting-section"):
|
| 408 |
-
# with gr.Row():
|
| 409 |
-
# available_variants = (
|
| 410 |
-
# list(all_model_cfg.keys()) if all_model_cfg else []
|
| 411 |
-
# )
|
| 412 |
-
# default_variant = (
|
| 413 |
-
# 'meanaudio_mf'
|
| 414 |
-
# )
|
| 415 |
-
# variant = gr.Dropdown(
|
| 416 |
-
# label="Model Variant",
|
| 417 |
-
# choices=available_variants,
|
| 418 |
-
# value=default_variant,
|
| 419 |
-
# interactive=True,
|
| 420 |
-
# scale=3,
|
| 421 |
-
# )
|
| 422 |
-
# with gr.Column(elem_classes="setting-section"):
|
| 423 |
-
# with gr.Row():
|
| 424 |
-
# prompt = gr.Textbox(
|
| 425 |
-
# label="Prompt",
|
| 426 |
-
# placeholder="Describe the sound you want to generate...",
|
| 427 |
-
# scale=1,
|
| 428 |
-
# )
|
| 429 |
-
# negative_prompt = gr.Textbox(
|
| 430 |
-
# label="Negative Prompt",
|
| 431 |
-
# placeholder="Describe sounds you want to avoid...",
|
| 432 |
-
# value="",
|
| 433 |
-
# scale=1,
|
| 434 |
-
# )
|
| 435 |
-
# with gr.Row():
|
| 436 |
-
# duration = gr.Number(
|
| 437 |
-
# label="Duration (sec)", value=10.0, minimum=0.1, scale=1
|
| 438 |
-
# )
|
| 439 |
-
# cfg_strength = gr.Number(
|
| 440 |
-
# label="CFG (Meanflow forced to 3)", value=3, minimum=0.0, scale=1
|
| 441 |
-
# )
|
| 442 |
-
# with gr.Row():
|
| 443 |
-
# seed = gr.Number(
|
| 444 |
-
# label="Seed (-1 for random)", value=42, precision=0, scale=1
|
| 445 |
-
# )
|
| 446 |
-
# num_steps = gr.Number(
|
| 447 |
-
# label="Number of Steps",
|
| 448 |
-
# value=1,
|
| 449 |
-
# precision=0,
|
| 450 |
-
# minimum=1,
|
| 451 |
-
# scale=1,
|
| 452 |
-
# )
|
| 453 |
-
# generate_button = gr.Button("Generate", variant="primary", elem_id="generate-btn")
|
| 454 |
-
# generate_output_text = gr.Textbox(
|
| 455 |
-
# label="Result Status", interactive=False, elem_id="status-box"
|
| 456 |
-
# )
|
| 457 |
-
# audio_output = gr.Audio(
|
| 458 |
-
# label="Generated Audio", type="filepath", elem_id="audio-output"
|
| 459 |
-
# )
|
| 460 |
-
# generate_button.click(
|
| 461 |
-
# fn=generate_audio_gradio,
|
| 462 |
-
# inputs=[
|
| 463 |
-
# prompt,
|
| 464 |
-
# negative_prompt,
|
| 465 |
-
# duration,
|
| 466 |
-
# cfg_strength,
|
| 467 |
-
# num_steps,
|
| 468 |
-
# seed,
|
| 469 |
-
# variant,
|
| 470 |
-
# ],
|
| 471 |
-
# outputs=[generate_output_text, audio_output],
|
| 472 |
-
# )
|
| 473 |
-
# audio_examples = [
|
| 474 |
-
# ["Typing on a keyboard", "", 10.0, 3, 1, 42, "meanaudio_mf"],
|
| 475 |
-
# ["A man speaks followed by a popping noise and laughter", "", 10.0, 3, 1, 42, "meanaudio_mf"],
|
| 476 |
-
# ["Some humming followed by a toilet flushing", "", 10.0, 3, 2, 42, "meanaudio_mf"],
|
| 477 |
-
# ["Rain falling on a hard surface as thunder roars in the distance", "", 10.0, 3, 5, 42, "meanaudio_mf"],
|
| 478 |
-
# ["Food sizzling and oil popping", "", 10.0, 3, 25, 42, "meanaudio_mf"],
|
| 479 |
-
# ["Pots and dishes clanking as a man talks followed by liquid pouring into a container", "", 8.0, 3, 2, 42, "meanaudio_mf"],
|
| 480 |
-
# ["A few seconds of silence then a rasping sound against wood", "", 12.0, 3, 2, 42, "meanaudio_mf"],
|
| 481 |
-
# ["A man speaks as he gives a speech and then the crowd cheers", "", 10.0, 3, 25, 42, "fluxaudio_fm"],
|
| 482 |
-
# ["A goat bleating repeatedly", "", 10.0, 3, 50, 123, "fluxaudio_fm"],
|
| 483 |
-
# ["A speech and gunfire followed by a gun being loaded", "", 10.0, 3, 1, 42, "meanaudio_mf"],
|
| 484 |
-
# ["Tires squealing followed by an engine revving", "", 12.0, 4, 25, 456, "fluxaudio_fm"],
|
| 485 |
-
# ["Hammer slowly hitting the wooden table", "", 10.0, 3.5, 25, 42, "fluxaudio_fm"],
|
| 486 |
-
# ["Dog barking excitedly and man shouting as race car engine roars past", "", 10.0, 3, 1, 42, "meanaudio_mf"],
|
| 487 |
-
# ["A dog barking and a cat mewing and a racing car passes by", "", 12.0, 3, 5, -1, "meanaudio_mf"],
|
| 488 |
-
# ["Whistling with birds chirping", "", 10.0, 4, 50, 42, "fluxaudio_fm"],
|
| 489 |
-
# ]
|
| 490 |
-
# gr.Examples(
|
| 491 |
-
# examples=audio_examples,
|
| 492 |
-
# inputs=[prompt, negative_prompt, duration, cfg_strength, num_steps, seed, variant],
|
| 493 |
-
# #outputs=[generate_output_text, audio_output],
|
| 494 |
-
# #fn=generate_audio_gradio,
|
| 495 |
-
# examples_per_page=5,
|
| 496 |
-
# label="Example Prompts",
|
| 497 |
-
# )
|
| 498 |
-
|
| 499 |
-
# if __name__ == "__main__":
|
| 500 |
-
# demo.launch()
|
|
|
|
| 243 |
|
| 244 |
|
| 245 |
if __name__ == "__main__":
|
| 246 |
+
ensure_models_downloaded()
|
| 247 |
+
load_model_cache()
|
| 248 |
+
gr_interface.queue(15).launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|