Update README.md
Browse files
README.md
CHANGED
|
@@ -1,578 +1,574 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
<
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
* Jun
|
| 29 |
-
[
|
| 30 |
-
* Jun
|
| 31 |
-
*
|
| 32 |
-
*
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
>
|
| 45 |
-
>
|
| 46 |
-
>
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
>
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
>
|
| 55 |
-
>
|
| 56 |
-
>
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
-
|
| 65 |
-
- [x]
|
| 66 |
-
- [x]
|
| 67 |
-
- [
|
| 68 |
-
- [
|
| 69 |
-
- [
|
| 70 |
-
- [
|
| 71 |
-
|
| 72 |
-
- [
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
- [
|
| 76 |
-
|
| 77 |
-
- [X]
|
| 78 |
-
- [
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
- [
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
- [
|
| 87 |
-
- [
|
| 88 |
-
- [
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
- [
|
| 92 |
-
- [
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
- [
|
| 97 |
-
- [
|
| 98 |
-
- [
|
| 99 |
-
- [
|
| 100 |
-
|
| 101 |
-
- [Using
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
| 133 |
-
|
| 134 |
-
|
| 135 |
-
<
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
<
|
| 139 |
-
|
| 140 |
-
</
|
| 141 |
-
</
|
| 142 |
-
<
|
| 143 |
-
<
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
<
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
<tr>
|
| 150 |
-
<td>
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
<tr>
|
| 154 |
-
<td>
|
| 155 |
-
|
| 156 |
-
|
| 157 |
-
<
|
| 158 |
-
|
| 159 |
-
|
| 160 |
-
<
|
| 161 |
-
|
| 162 |
-
</
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
|
| 166 |
-
|
| 167 |
-
|
| 168 |
-
|
| 169 |
-
|
| 170 |
-
|
| 171 |
-
|
| 172 |
-
|
| 173 |
-
|
| 174 |
-
|
| 175 |
-
|
| 176 |
-
|
| 177 |
-
|
| 178 |
-
|
| 179 |
-
|
| 180 |
-
|
| 181 |
-
|
| 182 |
-
|
| 183 |
-
|
| 184 |
-
|
| 185 |
-
|
| 186 |
-
|
| 187 |
-
|
| 188 |
-
|
| 189 |
-
|
| 190 |
-
|
| 191 |
-
|
| 192 |
-
|
| 193 |
-
|
| 194 |
-
|
| 195 |
-
|
| 196 |
-
|
| 197 |
-
|
| 198 |
-
|
|
| 199 |
-
|
| 200 |
-
|
| 201 |
-
|
| 202 |
-
|
| 203 |
-
|
| 204 |
-
|
| 205 |
-
*
|
| 206 |
-
|
| 207 |
-
|
| 208 |
-
|
| 209 |
-
|
| 210 |
-
|
| 211 |
-
|
| 212 |
-
|
| 213 |
-
|
| 214 |
-
|
| 215 |
-
|
| 216 |
-
|
| 217 |
-
|
| 218 |
-
|
| 219 |
-
|
| 220 |
-
|
| 221 |
-
|
| 222 |
-
|
| 223 |
-
|
| 224 |
-
|
| 225 |
-
|
| 226 |
-
|
| 227 |
-
|
| 228 |
-
|
| 229 |
-
|
| 230 |
-
|
| 231 |
-
|
| 232 |
-
|
| 233 |
-
|
| 234 |
-
|
| 235 |
-
|
| 236 |
-
|
| 237 |
-
|
| 238 |
-
|
| 239 |
-
|
| 240 |
-
|
| 241 |
-
|
| 242 |
-
|
| 243 |
-
```
|
| 244 |
-
|
| 245 |
-
|
| 246 |
-
|
| 247 |
-
|
| 248 |
-
|
| 249 |
-
|
| 250 |
-
|
| 251 |
-
|
| 252 |
-
huggingface-cli
|
| 253 |
-
|
| 254 |
-
|
| 255 |
-
|
| 256 |
-
|
| 257 |
-
|
| 258 |
-
|
| 259 |
-
|
| 260 |
-
|
| 261 |
-
|
| 262 |
-
|
| 263 |
-
|
| 264 |
-
|
| 265 |
-
|
| 266 |
-
|
| 267 |
-
|
| 268 |
-
|
| 269 |
-
|
| 270 |
-
|
| 271 |
-
|
| 272 |
-
|
| 273 |
-
|
| 274 |
-
|
| 275 |
-
|
| 276 |
-
|
| 277 |
-
|
| 278 |
-
|
| 279 |
-
|
| 280 |
-
|
|
| 281 |
-
|
| 282 |
-
|
|
| 283 |
-
|
|
| 284 |
-
|
|
| 285 |
-
|
| 286 |
-
|
| 287 |
-
|
| 288 |
-
|
| 289 |
-
|
| 290 |
-
|
| 291 |
-
|
| 292 |
-
|
| 293 |
-
|
| 294 |
-
|
| 295 |
-
|
| 296 |
-
|
| 297 |
-
|
| 298 |
-
|
| 299 |
-
|
| 300 |
-
|
| 301 |
-
|
| 302 |
-
|
| 303 |
-
```
|
| 304 |
-
|
| 305 |
-
|
| 306 |
-
|
| 307 |
-
|
| 308 |
-
|
| 309 |
-
|
| 310 |
-
|
| 311 |
-
|
| 312 |
-
|
| 313 |
-
|
| 314 |
-
|
| 315 |
-
|
| 316 |
-
|
| 317 |
-
|
|
| 318 |
-
|
| 319 |
-
| `
|
| 320 |
-
|
| 321 |
-
|
| 322 |
-
|
| 323 |
-
|
| 324 |
-
|
| 325 |
-
|
| 326 |
-
|
| 327 |
-
|
| 328 |
-
|
| 329 |
-
|
| 330 |
-
|
| 331 |
-
|
| 332 |
-
|
| 333 |
-
|
| 334 |
-
|
| 335 |
-
|
| 336 |
-
|
| 337 |
-
|
| 338 |
-
|
| 339 |
-
|
| 340 |
-
|
| 341 |
-
|
| 342 |
-
|
| 343 |
-
#
|
| 344 |
-
|
| 345 |
-
|
| 346 |
-
|
| 347 |
-
|
| 348 |
-
|
| 349 |
-
|
| 350 |
-
|
| 351 |
-
|
| 352 |
-
|
| 353 |
-
|
| 354 |
-
|
| 355 |
-
|
| 356 |
-
|
| 357 |
-
โโโ
|
| 358 |
-
โ โโโ
|
| 359 |
-
|
| 360 |
-
โ
|
| 361 |
-
โโโ
|
| 362 |
-
โ
|
| 363 |
-
โโโ
|
| 364 |
-
โ โโโ
|
| 365 |
-
โ โโโ
|
| 366 |
-
|
| 367 |
-
|
| 368 |
-
|
| 369 |
-
|
| 370 |
-
|
| 371 |
-
|
| 372 |
-
|
| 373 |
-
|
| 374 |
-
|
| 375 |
-
|
| 376 |
-
|
| 377 |
-
|
| 378 |
-
|
| 379 |
-
|
| 380 |
-
|
| 381 |
-
|
| 382 |
-
|
| 383 |
-
|
| 384 |
-
|
| 385 |
-
|
| 386 |
-
|
| 387 |
-
|
| 388 |
-
|
| 389 |
-
|
| 390 |
-
|
| 391 |
-
|
| 392 |
-
|
| 393 |
-
|
| 394 |
-
|
| 395 |
-
|
| 396 |
-
|
| 397 |
-
|
| 398 |
-
|
| 399 |
-
#
|
| 400 |
-
python app/hydit_app.py --
|
| 401 |
-
|
| 402 |
-
#
|
| 403 |
-
|
| 404 |
-
|
| 405 |
-
|
| 406 |
-
#
|
| 407 |
-
python app/
|
| 408 |
-
|
| 409 |
-
|
| 410 |
-
|
| 411 |
-
|
| 412 |
-
|
| 413 |
-
|
| 414 |
-
|
| 415 |
-
|
| 416 |
-
|
| 417 |
-
|
| 418 |
-
|
| 419 |
-
|
| 420 |
-
|
| 421 |
-
```shell
|
| 422 |
-
pip install
|
| 423 |
-
```
|
| 424 |
-
|
| 425 |
-
|
| 426 |
-
|
| 427 |
-
|
| 428 |
-
|
| 429 |
-
|
| 430 |
-
|
| 431 |
-
|
| 432 |
-
|
| 433 |
-
|
| 434 |
-
|
| 435 |
-
|
| 436 |
-
|
| 437 |
-
|
| 438 |
-
|
| 439 |
-
|
| 440 |
-
|
| 441 |
-
|
| 442 |
-
|
| 443 |
-
|
| 444 |
-
|
| 445 |
-
|
| 446 |
-
|
| 447 |
-
|
| 448 |
-
|
| 449 |
-
|
| 450 |
-
|
| 451 |
-
|
| 452 |
-
|
| 453 |
-
|
| 454 |
-
|
| 455 |
-
|
| 456 |
-
|
| 457 |
-
|
| 458 |
-
|
| 459 |
-
|
| 460 |
-
|
| 461 |
-
|
| 462 |
-
|
| 463 |
-
|
| 464 |
-
|
| 465 |
-
|
| 466 |
-
|
| 467 |
-
|
| 468 |
-
|
| 469 |
-
|
| 470 |
-
|
| 471 |
-
|
| 472 |
-
|
| 473 |
-
|
| 474 |
-
|
| 475 |
-
|
| 476 |
-
|
| 477 |
-
|
| 478 |
-
|
| 479 |
-
|
| 480 |
-
|
| 481 |
-
|
| 482 |
-
|
| 483 |
-
|
| 484 |
-
|
| 485 |
-
|
| 486 |
-
|
|
| 487 |
-
|
| 488 |
-
|
|
| 489 |
-
| `--
|
| 490 |
-
|
|
| 491 |
-
| `--
|
| 492 |
-
|
|
| 493 |
-
|
|
| 494 |
-
|
|
| 495 |
-
|
| 496 |
-
|
| 497 |
-
|
| 498 |
-
|
| 499 |
-
|
| 500 |
-
|
| 501 |
-
|
| 502 |
-
|
| 503 |
-
|
| 504 |
-
|
| 505 |
-
|
| 506 |
-
|
| 507 |
-
|
| 508 |
-
|
| 509 |
-
pip install
|
| 510 |
-
|
| 511 |
-
#
|
| 512 |
-
|
| 513 |
-
|
| 514 |
-
|
| 515 |
-
|
| 516 |
-
|
| 517 |
-
|
| 518 |
-
|
| 519 |
-
|
| 520 |
-
|
| 521 |
-
|
| 522 |
-
|
| 523 |
-
|
| 524 |
-
|
| 525 |
-
|
| 526 |
-
|
| 527 |
-
|
| 528 |
-
|
| 529 |
-
|
| 530 |
-
#
|
| 531 |
-
|
| 532 |
-
|
| 533 |
-
|
| 534 |
-
|
| 535 |
-
|
| 536 |
-
|
| 537 |
-
|
| 538 |
-
|
| 539 |
-
|
| 540 |
-
|
| 541 |
-
|
| 542 |
-
|
| 543 |
-
-
|
| 544 |
-
|
| 545 |
-
|
| 546 |
-
|
| 547 |
-
|
| 548 |
-
|
| 549 |
-
|
| 550 |
-
|
| 551 |
-
|
| 552 |
-
|
| 553 |
-
|
| 554 |
-
|
| 555 |
-
|
| 556 |
-
|
| 557 |
-
|
| 558 |
-
|
| 559 |
-
|
| 560 |
-
}
|
| 561 |
-
|
| 562 |
-
|
| 563 |
-
|
| 564 |
-
|
| 565 |
-
|
| 566 |
-
|
| 567 |
-
|
| 568 |
-
|
| 569 |
-
|
| 570 |
-
|
| 571 |
-
|
| 572 |
-
<
|
| 573 |
-
|
| 574 |
-
|
| 575 |
-
<source media="(prefers-color-scheme: light)" srcset="https://api.star-history.com/svg?repos=Tencent/HunyuanDiT&type=Date" />
|
| 576 |
-
<img alt="Star History Chart" src="https://api.star-history.com/svg?repos=Tencent/HunyuanDiT&type=Date" />
|
| 577 |
-
</picture>
|
| 578 |
-
</a>
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
license_name: tencent-hunyuan-community
|
| 4 |
+
license_link: https://huggingface.co/Tencent-Hunyuan/HunyuanDiT/blob/main/LICENSE.txt
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
---
|
| 8 |
+
<!-- ## **HunyuanDiT** -->
|
| 9 |
+
|
| 10 |
+
<p align="center">
|
| 11 |
+
<img src="https://raw.githubusercontent.com/Tencent/HunyuanDiT/main/asset/logo.png" height=100>
|
| 12 |
+
</p>
|
| 13 |
+
|
| 14 |
+
# Hunyuan-DiT : A Powerful Multi-Resolution Diffusion Transformer with Fine-Grained Chinese Understanding
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
This repo contains PyTorch model definitions, pre-trained weights and inference/sampling code for our paper exploring Hunyuan-DiT. You can find more visualizations on our [project page](https://dit.hunyuan.tencent.com/).
|
| 18 |
+
|
| 19 |
+
> [**Hunyuan-DiT: A Powerful Multi-Resolution Diffusion Transformer with Fine-Grained Chinese Understanding**](https://arxiv.org/abs/2405.08748) <br>
|
| 20 |
+
|
| 21 |
+
> [**DialogGen: Multi-modal Interactive Dialogue System for Multi-turn Text-to-Image Generation**](https://arxiv.org/abs/2403.08857) <br>
|
| 22 |
+
|
| 23 |
+
## ๐ฅ๐ฅ๐ฅ News!!
|
| 24 |
+
* Jun 13, 2024: :zap: HYDiT-v1.1 version is released, which mitigates the issue of image oversaturation and alleviates the watermark issue. Please check [HunyuanDiT-v1.1 ](https://huggingface.co/Tencent-Hunyuan/HunyuanDiT-v1.1) and
|
| 25 |
+
[Distillation-v1.1](https://huggingface.co/Tencent-Hunyuan/Distillation-v1.1) for more details.
|
| 26 |
+
* Jun 13, 2024: :truck: The training code is released, offering [full-parameter training](#full-parameter-training) and [LoRA training](#lora).
|
| 27 |
+
* Jun 06, 2024: :tada: Hunyuan-DiT is now available in ComfyUI. Please check [ComfyUI](#using-comfyui) for more details.
|
| 28 |
+
* Jun 06, 2024: ๐ We introduce Distillation version for Hunyuan-DiT acceleration, which achieves **50%** acceleration on NVIDIA GPUs. Please check [Distillation](https://huggingface.co/Tencent-Hunyuan/Distillation) for more details.
|
| 29 |
+
* Jun 05, 2024: ๐ค Hunyuan-DiT is now available in ๐ค Diffusers! Please check the [example](#using--diffusers) below.
|
| 30 |
+
* Jun 04, 2024: :globe_with_meridians: Support Tencent Cloud links to download the pretrained models! Please check the [links](#-download-pretrained-models) below.
|
| 31 |
+
* May 22, 2024: ๐ We introduce TensorRT version for Hunyuan-DiT acceleration, which achieves **47%** acceleration on NVIDIA GPUs. Please check [TensorRT-libs](https://huggingface.co/Tencent-Hunyuan/TensorRT-libs) for instructions.
|
| 32 |
+
* May 22, 2024: ๐ฌ We support demo running multi-turn text2image generation now. Please check the [script](#using-gradio) below.
|
| 33 |
+
|
| 34 |
+
## ๐ค Try it on the web
|
| 35 |
+
|
| 36 |
+
Welcome to our web-based [**Tencent Hunyuan Bot**](https://hunyuan.tencent.com/bot/chat), where you can explore our innovative products! Just input the suggested prompts below or any other **imaginative prompts containing drawing-related keywords** to activate the Hunyuan text-to-image generation feature. Unleash your creativity and create any picture you desire, **all for free!**
|
| 37 |
+
|
| 38 |
+
You can use simple prompts similar to natural language text
|
| 39 |
+
|
| 40 |
+
> ็ปไธๅช็ฉฟ็่ฅฟ่ฃ
็็ช
|
| 41 |
+
>
|
| 42 |
+
> draw a pig in a suit
|
| 43 |
+
>
|
| 44 |
+
> ็ๆไธๅน
็ป๏ผ่ตๅๆๅ
้ฃ๏ผ่ท่ฝฆ
|
| 45 |
+
>
|
| 46 |
+
> generate a painting, cyberpunk style, sports car
|
| 47 |
+
|
| 48 |
+
or multi-turn language interactions to create the picture.
|
| 49 |
+
|
| 50 |
+
> ็ปไธไธชๆจๅถ็้ธ
|
| 51 |
+
>
|
| 52 |
+
> draw a wooden bird
|
| 53 |
+
>
|
| 54 |
+
> ๅๆ็ป็็
|
| 55 |
+
>
|
| 56 |
+
> turn into glass
|
| 57 |
+
|
| 58 |
+
## ๐ Open-source Plan
|
| 59 |
+
|
| 60 |
+
- Hunyuan-DiT (Text-to-Image Model)
|
| 61 |
+
- [x] Inference
|
| 62 |
+
- [x] Checkpoints
|
| 63 |
+
- [x] Distillation Version
|
| 64 |
+
- [x] TensorRT Version
|
| 65 |
+
- [x] Training
|
| 66 |
+
- [x] Lora
|
| 67 |
+
- [ ] Controlnet (Pose, Canny, Depth, Tile)
|
| 68 |
+
- [ ] IP-adapter
|
| 69 |
+
- [ ] Hunyuan-DiT-XL checkpoints (0.7B model)
|
| 70 |
+
- [ ] Caption model (Re-caption the raw image-text pairs)
|
| 71 |
+
- [DialogGen](https://github.com/Centaurusalpha/DialogGen) (Prompt Enhancement Model)
|
| 72 |
+
- [x] Inference
|
| 73 |
+
- [X] Web Demo (Gradio)
|
| 74 |
+
- [x] Multi-turn T2I Demo (Gradio)
|
| 75 |
+
- [X] Cli Demo
|
| 76 |
+
- [X] ComfyUI
|
| 77 |
+
- [X] Diffusers
|
| 78 |
+
- [ ] WebUI
|
| 79 |
+
|
| 80 |
+
|
| 81 |
+
## Contents
|
| 82 |
+
- [Hunyuan-DiT](#hunyuan-dit--a-powerful-multi-resolution-diffusion-transformer-with-fine-grained-chinese-understanding)
|
| 83 |
+
- [Abstract](#abstract)
|
| 84 |
+
- [๐ Hunyuan-DiT Key Features](#-hunyuan-dit-key-features)
|
| 85 |
+
- [Chinese-English Bilingual DiT Architecture](#chinese-english-bilingual-dit-architecture)
|
| 86 |
+
- [Multi-turn Text2Image Generation](#multi-turn-text2image-generation)
|
| 87 |
+
- [๐ Comparisons](#-comparisons)
|
| 88 |
+
- [๐ฅ Visualization](#-visualization)
|
| 89 |
+
- [๐ Requirements](#-requirements)
|
| 90 |
+
- [๐ Dependencies and Installation](#%EF%B8%8F-dependencies-and-installation)
|
| 91 |
+
- [๐งฑ Download Pretrained Models](#-download-pretrained-models)
|
| 92 |
+
- [:truck: Training](#truck-training)
|
| 93 |
+
- [Data Preparation](#data-preparation)
|
| 94 |
+
- [Full Parameter Training](#full-parameter-training)
|
| 95 |
+
- [LoRA](#lora)
|
| 96 |
+
- [๐ Inference](#-inference)
|
| 97 |
+
- [Using Gradio](#using-gradio)
|
| 98 |
+
- [Using Diffusers](#using--diffusers)
|
| 99 |
+
- [Using Command Line](#using-command-line)
|
| 100 |
+
- [More Configurations](#more-configurations)
|
| 101 |
+
- [Using ComfyUI](#using-comfyui)
|
| 102 |
+
- [๐ Acceleration (for Linux)](#-acceleration-for-linux)
|
| 103 |
+
- [๐ BibTeX](#-bibtex)
|
| 104 |
+
|
| 105 |
+
## **Abstract**
|
| 106 |
+
|
| 107 |
+
We present Hunyuan-DiT, a text-to-image diffusion transformer with fine-grained understanding of both English and Chinese. To construct Hunyuan-DiT, we carefully designed the transformer structure, text encoder, and positional encoding. We also build from scratch a whole data pipeline to update and evaluate data for iterative model optimization. For fine-grained language understanding, we train a Multimodal Large Language Model to refine the captions of the images. Finally, Hunyuan-DiT can perform multi-round multi-modal dialogue with users, generating and refining images according to the context.
|
| 108 |
+
Through our carefully designed holistic human evaluation protocol with more than 50 professional human evaluators, Hunyuan-DiT sets a new state-of-the-art in Chinese-to-image generation compared with other open-source models.
|
| 109 |
+
|
| 110 |
+
|
| 111 |
+
## ๐ **Hunyuan-DiT Key Features**
|
| 112 |
+
### **Chinese-English Bilingual DiT Architecture**
|
| 113 |
+
Hunyuan-DiT is a diffusion model in the latent space, as depicted in figure below. Following the Latent Diffusion Model, we use a pre-trained Variational Autoencoder (VAE) to compress the images into low-dimensional latent spaces and train a diffusion model to learn the data distribution with diffusion models. Our diffusion model is parameterized with a transformer. To encode the text prompts, we leverage a combination of pre-trained bilingual (English and Chinese) CLIP and multilingual T5 encoder.
|
| 114 |
+
<p align="center">
|
| 115 |
+
<img src="https://raw.githubusercontent.com/Tencent/HunyuanDiT/main/asset/framework.png" height=450>
|
| 116 |
+
</p>
|
| 117 |
+
|
| 118 |
+
### Multi-turn Text2Image Generation
|
| 119 |
+
Understanding natural language instructions and performing multi-turn interaction with users are important for a
|
| 120 |
+
text-to-image system. It can help build a dynamic and iterative creation process that bring the userโs idea into reality
|
| 121 |
+
step by step. In this section, we will detail how we empower Hunyuan-DiT with the ability to perform multi-round
|
| 122 |
+
conversations and image generation. We train MLLM to understand the multi-round user dialogue
|
| 123 |
+
and output the new text prompt for image generation.
|
| 124 |
+
<p align="center">
|
| 125 |
+
<img src="https://raw.githubusercontent.com/Tencent/HunyuanDiT/main/asset/mllm.png" height=300>
|
| 126 |
+
</p>
|
| 127 |
+
|
| 128 |
+
## ๐ Comparisons
|
| 129 |
+
In order to comprehensively compare the generation capabilities of HunyuanDiT and other models, we constructed a 4-dimensional test set, including Text-Image Consistency, Excluding AI Artifacts, Subject Clarity, Aesthetic. More than 50 professional evaluators performs the evaluation.
|
| 130 |
+
|
| 131 |
+
<p align="center">
|
| 132 |
+
<table>
|
| 133 |
+
<thead>
|
| 134 |
+
<tr>
|
| 135 |
+
<th rowspan="2">Model</th> <th rowspan="2">Open Source</th> <th>Text-Image Consistency (%)</th> <th>Excluding AI Artifacts (%)</th> <th>Subject Clarity (%)</th> <th rowspan="2">Aesthetics (%)</th> <th rowspan="2">Overall (%)</th>
|
| 136 |
+
</tr>
|
| 137 |
+
</thead>
|
| 138 |
+
<tbody>
|
| 139 |
+
<tr>
|
| 140 |
+
<td>SDXL</td> <td> โ </td> <td>64.3</td> <td>60.6</td> <td>91.1</td> <td>76.3</td> <td>42.7</td>
|
| 141 |
+
</tr>
|
| 142 |
+
<tr>
|
| 143 |
+
<td>PixArt-ฮฑ</td> <td> โ</td> <td>68.3</td> <td>60.9</td> <td>93.2</td> <td>77.5</td> <td>45.5</td>
|
| 144 |
+
</tr>
|
| 145 |
+
<tr>
|
| 146 |
+
<td>Playground 2.5</td> <td>โ</td> <td>71.9</td> <td>70.8</td> <td>94.9</td> <td>83.3</td> <td>54.3</td>
|
| 147 |
+
</tr>
|
| 148 |
+
|
| 149 |
+
<tr>
|
| 150 |
+
<td>SD 3</td> <td>✘</td> <td>77.1</td> <td>69.3</td> <td>94.6</td> <td>82.5</td> <td>56.7</td>
|
| 151 |
+
|
| 152 |
+
</tr>
|
| 153 |
+
<tr>
|
| 154 |
+
<td>MidJourney v6</td><td>✘</td> <td>73.5</td> <td>80.2</td> <td>93.5</td> <td>87.2</td> <td>63.3</td>
|
| 155 |
+
</tr>
|
| 156 |
+
<tr>
|
| 157 |
+
<td>DALL-E 3</td><td>✘</td> <td>83.9</td> <td>80.3</td> <td>96.5</td> <td>89.4</td> <td>71.0</td>
|
| 158 |
+
</tr>
|
| 159 |
+
<tr style="font-weight: bold; background-color: #f2f2f2;">
|
| 160 |
+
<td>Hunyuan-DiT</td><td>โ</td> <td>74.2</td> <td>74.3</td> <td>95.4</td> <td>86.6</td> <td>59.0</td>
|
| 161 |
+
</tr>
|
| 162 |
+
</tbody>
|
| 163 |
+
</table>
|
| 164 |
+
</p>
|
| 165 |
+
|
| 166 |
+
## ๐ฅ Visualization
|
| 167 |
+
|
| 168 |
+
* **Chinese Elements**
|
| 169 |
+
<p align="center">
|
| 170 |
+
<img src="https://raw.githubusercontent.com/Tencent/HunyuanDiT/main/asset/chinese elements understanding.png" height=220>
|
| 171 |
+
</p>
|
| 172 |
+
|
| 173 |
+
* **Long Text Input**
|
| 174 |
+
|
| 175 |
+
|
| 176 |
+
<p align="center">
|
| 177 |
+
<img src="https://raw.githubusercontent.com/Tencent/HunyuanDiT/main/asset/long text understanding.png" height=310>
|
| 178 |
+
</p>
|
| 179 |
+
|
| 180 |
+
* **Multi-turn Text2Image Generation**
|
| 181 |
+
|
| 182 |
+
https://github.com/Tencent/tencent.github.io/assets/27557933/94b4dcc3-104d-44e1-8bb2-dc55108763d1
|
| 183 |
+
|
| 184 |
+
|
| 185 |
+
|
| 186 |
+
---
|
| 187 |
+
|
| 188 |
+
## ๐ Requirements
|
| 189 |
+
|
| 190 |
+
This repo consists of DialogGen (a prompt enhancement model) and Hunyuan-DiT (a text-to-image model).
|
| 191 |
+
|
| 192 |
+
The following table shows the requirements for running the models (batch size = 1):
|
| 193 |
+
|
| 194 |
+
| Model | --load-4bit (DialogGen) | GPU Peak Memory | GPU |
|
| 195 |
+
|:-----------------------:|:-----------------------:|:---------------:|:---------------:|
|
| 196 |
+
| DialogGen + Hunyuan-DiT | โ | 32G | A100 |
|
| 197 |
+
| DialogGen + Hunyuan-DiT | โ | 22G | A100 |
|
| 198 |
+
| Hunyuan-DiT | - | 11G | A100 |
|
| 199 |
+
| Hunyuan-DiT | - | 14G | RTX3090/RTX4090 |
|
| 200 |
+
|
| 201 |
+
* An NVIDIA GPU with CUDA support is required.
|
| 202 |
+
* We have tested V100 and A100 GPUs.
|
| 203 |
+
* **Minimum**: The minimum GPU memory required is 11GB.
|
| 204 |
+
* **Recommended**: We recommend using a GPU with 32GB of memory for better generation quality.
|
| 205 |
+
* Tested operating system: Linux
|
| 206 |
+
|
| 207 |
+
## ๐ ๏ธ Dependencies and Installation
|
| 208 |
+
|
| 209 |
+
Begin by cloning the repository:
|
| 210 |
+
```shell
|
| 211 |
+
git clone https://github.com/tencent/HunyuanDiT
|
| 212 |
+
cd HunyuanDiT
|
| 213 |
+
```
|
| 214 |
+
|
| 215 |
+
### Installation Guide for Linux
|
| 216 |
+
|
| 217 |
+
We provide an `environment.yml` file for setting up a Conda environment.
|
| 218 |
+
Conda's installation instructions are available [here](https://docs.anaconda.com/free/miniconda/index.html).
|
| 219 |
+
|
| 220 |
+
```shell
|
| 221 |
+
# 1. Prepare conda environment
|
| 222 |
+
conda env create -f environment.yml
|
| 223 |
+
|
| 224 |
+
# 2. Activate the environment
|
| 225 |
+
conda activate HunyuanDiT
|
| 226 |
+
|
| 227 |
+
# 3. Install pip dependencies
|
| 228 |
+
python -m pip install -r requirements.txt
|
| 229 |
+
|
| 230 |
+
# 4. (Optional) Install flash attention v2 for acceleration (requires CUDA 11.6 or above)
|
| 231 |
+
python -m pip install git+https://github.com/Dao-AILab/[email protected]
|
| 232 |
+
```
|
| 233 |
+
|
| 234 |
+
## ๐งฑ Download Pretrained Models
|
| 235 |
+
To download the model, first install the huggingface-cli. (Detailed instructions are available [here](https://huggingface.co/docs/huggingface_hub/guides/cli).)
|
| 236 |
+
|
| 237 |
+
```shell
|
| 238 |
+
python -m pip install "huggingface_hub[cli]"
|
| 239 |
+
```
|
| 240 |
+
|
| 241 |
+
Then download the model using the following commands:
|
| 242 |
+
|
| 243 |
+
```shell
|
| 244 |
+
# Create a directory named 'ckpts' where the model will be saved, fulfilling the prerequisites for running the demo.
|
| 245 |
+
mkdir ckpts
|
| 246 |
+
# Use the huggingface-cli tool to download the model.
|
| 247 |
+
# The download time may vary from 10 minutes to 1 hour depending on network conditions.
|
| 248 |
+
huggingface-cli download Tencent-Hunyuan/HunyuanDiT --local-dir ./ckpts
|
| 249 |
+
```
|
| 250 |
+
|
| 251 |
+
<details>
|
| 252 |
+
<summary>๐กTips for using huggingface-cli (network problem)</summary>
|
| 253 |
+
|
| 254 |
+
##### 1. Using HF-Mirror
|
| 255 |
+
|
| 256 |
+
If you encounter slow download speeds in China, you can try a mirror to speed up the download process. For example,
|
| 257 |
+
|
| 258 |
+
```shell
|
| 259 |
+
HF_ENDPOINT=https://hf-mirror.com huggingface-cli download Tencent-Hunyuan/HunyuanDiT --local-dir ./ckpts
|
| 260 |
+
```
|
| 261 |
+
|
| 262 |
+
##### 2. Resume Download
|
| 263 |
+
|
| 264 |
+
`huggingface-cli` supports resuming downloads. If the download is interrupted, you can just rerun the download
|
| 265 |
+
command to resume the download process.
|
| 266 |
+
|
| 267 |
+
Note: If an `No such file or directory: 'ckpts/.huggingface/.gitignore.lock'` like error occurs during the download
|
| 268 |
+
process, you can ignore the error and rerun the download command.
|
| 269 |
+
|
| 270 |
+
</details>
|
| 271 |
+
|
| 272 |
+
---
|
| 273 |
+
|
| 274 |
+
All models will be automatically downloaded. For more information about the model, visit the Hugging Face repository [here](https://huggingface.co/Tencent-Hunyuan/HunyuanDiT).
|
| 275 |
+
|
| 276 |
+
| Model | #Params | Huggingface Download URL | Tencent Cloud Download URL |
|
| 277 |
+
|:------------------:|:-------:|:-------------------------------------------------------------------------------------------------------:|:-----------------------------------------------------------------------------------------------:|
|
| 278 |
+
| mT5 | 1.6B | [mT5](https://huggingface.co/Tencent-Hunyuan/HunyuanDiT/tree/main/t2i/mt5) | [mT5](https://dit.hunyuan.tencent.com/download/HunyuanDiT/mt5.zip) |
|
| 279 |
+
| CLIP | 350M | [CLIP](https://huggingface.co/Tencent-Hunyuan/HunyuanDiT/tree/main/t2i/clip_text_encoder) | [CLIP](https://dit.hunyuan.tencent.com/download/HunyuanDiT/clip_text_encoder.zip) |
|
| 280 |
+
| Tokenizer | - | [Tokenizer](https://huggingface.co/Tencent-Hunyuan/HunyuanDiT/tree/main/t2i/tokenizer) | [Tokenizer](https://dit.hunyuan.tencent.com/download/HunyuanDiT/tokenizer.zip) |
|
| 281 |
+
| DialogGen | 7.0B | [DialogGen](https://huggingface.co/Tencent-Hunyuan/HunyuanDiT/tree/main/dialoggen) | [DialogGen](https://dit.hunyuan.tencent.com/download/HunyuanDiT/dialoggen.zip) |
|
| 282 |
+
| sdxl-vae-fp16-fix | 83M | [sdxl-vae-fp16-fix](https://huggingface.co/Tencent-Hunyuan/HunyuanDiT/tree/main/t2i/sdxl-vae-fp16-fix) | [sdxl-vae-fp16-fix](https://dit.hunyuan.tencent.com/download/HunyuanDiT/sdxl-vae-fp16-fix.zip) |
|
| 283 |
+
| Hunyuan-DiT | 1.5B | [Hunyuan-DiT](https://huggingface.co/Tencent-Hunyuan/HunyuanDiT/tree/main/t2i/model) | [Hunyuan-DiT](https://dit.hunyuan.tencent.com/download/HunyuanDiT/model.zip) |
|
| 284 |
+
| Data demo | - | - | [Data demo](https://dit.hunyuan.tencent.com/download/HunyuanDiT/data_demo.zip) |
|
| 285 |
+
|
| 286 |
+
## :truck: Training
|
| 287 |
+
|
| 288 |
+
### Data Preparation
|
| 289 |
+
|
| 290 |
+
Refer to the commands below to prepare the training data.
|
| 291 |
+
|
| 292 |
+
1. Install dependencies
|
| 293 |
+
|
| 294 |
+
We offer an efficient data management library, named IndexKits, supporting the management of reading hundreds of millions of data during training, see more in [docs](./IndexKits/README.md).
|
| 295 |
+
```shell
|
| 296 |
+
# 1 Install dependencies
|
| 297 |
+
cd HunyuanDiT
|
| 298 |
+
pip install -e ./IndexKits
|
| 299 |
+
```
|
| 300 |
+
2. Data download
|
| 301 |
+
|
| 302 |
+
Feel free to download the [data demo](https://dit.hunyuan.tencent.com/download/HunyuanDiT/data_demo.zip).
|
| 303 |
+
```shell
|
| 304 |
+
# 2 Data download
|
| 305 |
+
wget -O ./dataset/data_demo.zip https://dit.hunyuan.tencent.com/download/HunyuanDiT/data_demo.zip
|
| 306 |
+
unzip ./dataset/data_demo.zip -d ./dataset
|
| 307 |
+
mkdir ./dataset/porcelain/arrows ./dataset/porcelain/jsons
|
| 308 |
+
```
|
| 309 |
+
3. Data conversion
|
| 310 |
+
|
| 311 |
+
Create a CSV file for training data with the fields listed in the table below.
|
| 312 |
+
|
| 313 |
+
| Fields | Required | Description | Example |
|
| 314 |
+
|:---------------:| :------: |:----------------:|:-----------:|
|
| 315 |
+
| `image_path` | Required | image path | `./dataset/porcelain/images/0.png` |
|
| 316 |
+
| `text_zh` | Required | text | ้่ฑ็ท้ฃๆ ผ๏ผไธๅช่่ฒ็้ธๅฟ็ซๅจ่่ฒ็่ฑ็ถไธ๏ผๅจๅด็น็ผ็็ฝ่ฒ่ฑๆต๏ผ่ๆฏๆฏ็ฝ่ฒ |
|
| 317 |
+
| `md5` | Optional | image md5 (Message Digest Algorithm 5) | `d41d8cd98f00b204e9800998ecf8427e` |
|
| 318 |
+
| `width` | Optional | image width | `1024 ` |
|
| 319 |
+
| `height` | Optional | image height | ` 1024 ` |
|
| 320 |
+
|
| 321 |
+
> โ ๏ธ Optional fields like MD5, width, and height can be omitted. If omitted, the script below will automatically calculate them. This process can be time-consuming when dealing with large-scale training data.
|
| 322 |
+
|
| 323 |
+
We utilize [Arrow](https://github.com/apache/arrow) for training data format, offering a standard and efficient in-memory data representation. A conversion script is provided to transform CSV files into Arrow format.
|
| 324 |
+
```shell
|
| 325 |
+
# 3 Data conversion
|
| 326 |
+
python ./hydit/data_loader/csv2arrow.py ./dataset/porcelain/csvfile/image_text.csv ./dataset/porcelain/arrows
|
| 327 |
+
```
|
| 328 |
+
|
| 329 |
+
4. Data Selection and Configuration File Creation
|
| 330 |
+
|
| 331 |
+
We configure the training data through YAML files. In these files, you can set up standard data processing strategies for filtering, copying, deduplicating, and more regarding the training data. For more details, see [docs](IndexKits/docs/MakeDataset.md).
|
| 332 |
+
|
| 333 |
+
For a sample file, please refer to [file](./dataset/yamls/porcelain.yaml). For a full parameter configuration file, see [file](./IndexKits/docs/MakeDataset.md).
|
| 334 |
+
|
| 335 |
+
|
| 336 |
+
5. Create training data index file using YAML file.
|
| 337 |
+
|
| 338 |
+
```shell
|
| 339 |
+
# Single Resolution Data Preparation
|
| 340 |
+
cd /HunyuanDiT
|
| 341 |
+
idk base -c dataset/yamls/porcelain.yaml -t dataset/porcelain/jsons/porcelain.json
|
| 342 |
+
|
| 343 |
+
# Multi Resolution Data Preparation
|
| 344 |
+
idk multireso -c dataset/yamls/porcelain_mt.yaml -t dataset/porcelain/jsons/porcelain_mt.json
|
| 345 |
+
```
|
| 346 |
+
|
| 347 |
+
The directory structure for `porcelain` dataset is:
|
| 348 |
+
|
| 349 |
+
```shell
|
| 350 |
+
cd ./dataset
|
| 351 |
+
|
| 352 |
+
porcelain
|
| 353 |
+
โโโimages/ (image files)
|
| 354 |
+
โ โโโ0.png
|
| 355 |
+
โ โโโ1.png
|
| 356 |
+
โ โโโ......
|
| 357 |
+
โโโcsvfile/ (csv files containing text-image pairs)
|
| 358 |
+
โ โโโimage_text.csv
|
| 359 |
+
โโโarrows/ (arrow files containing all necessary training data)
|
| 360 |
+
โ โโโ00000.arrow
|
| 361 |
+
โ โโโ00001.arrow
|
| 362 |
+
โ โโโ......
|
| 363 |
+
โโโjsons/ (final training data index files which read data from arrow files during training)
|
| 364 |
+
โ โโโporcelain.json
|
| 365 |
+
โ โโโporcelain_mt.json
|
| 366 |
+
```
|
| 367 |
+
|
| 368 |
+
### Full-parameter Training
|
| 369 |
+
|
| 370 |
+
To leverage DeepSpeed in training, you have the flexibility to control **single-node** / **multi-node** training by adjusting parameters such as `--hostfile` and `--master_addr`. For more details, see [link](https://www.deepspeed.ai/getting-started/#resource-configuration-multi-node).
|
| 371 |
+
|
| 372 |
+
```shell
|
| 373 |
+
# Single Resolution Data Preparation
|
| 374 |
+
PYTHONPATH=./ sh hydit/train.sh --index-file dataset/porcelain/jsons/porcelain.json
|
| 375 |
+
|
| 376 |
+
# Multi Resolution Data Preparation
|
| 377 |
+
PYTHONPATH=./ sh hydit/train.sh --index-file dataset/porcelain/jsons/porcelain.json --multireso --reso-step 64
|
| 378 |
+
```
|
| 379 |
+
|
| 380 |
+
### LoRA
|
| 381 |
+
|
| 382 |
+
We provide training and inference scripts for LoRA, detailed in the [guidances](./lora/README.md).
|
| 383 |
+
|
| 384 |
+
|
| 385 |
+
## ๐ Inference
|
| 386 |
+
|
| 387 |
+
### Using Gradio
|
| 388 |
+
|
| 389 |
+
Make sure the conda environment is activated before running the following command.
|
| 390 |
+
|
| 391 |
+
```shell
|
| 392 |
+
# By default, we start a Chinese UI.
|
| 393 |
+
python app/hydit_app.py
|
| 394 |
+
|
| 395 |
+
# Using Flash Attention for acceleration.
|
| 396 |
+
python app/hydit_app.py --infer-mode fa
|
| 397 |
+
|
| 398 |
+
# You can disable the enhancement model if the GPU memory is insufficient.
|
| 399 |
+
# The enhancement will be unavailable until you restart the app without the `--no-enhance` flag.
|
| 400 |
+
python app/hydit_app.py --no-enhance
|
| 401 |
+
|
| 402 |
+
# Start with English UI
|
| 403 |
+
python app/hydit_app.py --lang en
|
| 404 |
+
|
| 405 |
+
# Start a multi-turn T2I generation UI.
|
| 406 |
+
# If your GPU memory is less than 32GB, use '--load-4bit' to enable 4-bit quantization, which requires at least 22GB of memory.
|
| 407 |
+
python app/multiTurnT2I_app.py
|
| 408 |
+
```
|
| 409 |
+
Then the demo can be accessed through http://0.0.0.0:443. It should be noted that the 0.0.0.0 here needs to be X.X.X.X with your server IP.
|
| 410 |
+
|
| 411 |
+
### Using ๐ค Diffusers
|
| 412 |
+
|
| 413 |
+
Please install PyTorch version 2.0 or higher in advance to satisfy the requirements of the specified version of the diffusers library.
|
| 414 |
+
|
| 415 |
+
Install ๐ค diffusers, ensuring that the version is at least 0.28.1:
|
| 416 |
+
|
| 417 |
+
```shell
|
| 418 |
+
pip install git+https://github.com/huggingface/diffusers.git
|
| 419 |
+
```
|
| 420 |
+
or
|
| 421 |
+
```shell
|
| 422 |
+
pip install diffusers
|
| 423 |
+
```
|
| 424 |
+
|
| 425 |
+
You can generate images with both Chinese and English prompts using the following Python script:
|
| 426 |
+
```py
|
| 427 |
+
import torch
|
| 428 |
+
from diffusers import HunyuanDiTPipeline
|
| 429 |
+
|
| 430 |
+
pipe = HunyuanDiTPipeline.from_pretrained("Tencent-Hunyuan/HunyuanDiT-Diffusers", torch_dtype=torch.float16)
|
| 431 |
+
pipe.to("cuda")
|
| 432 |
+
|
| 433 |
+
# You may also use English prompt as HunyuanDiT supports both English and Chinese
|
| 434 |
+
# prompt = "An astronaut riding a horse"
|
| 435 |
+
prompt = "ไธไธชๅฎ่ชๅๅจ้ช้ฉฌ"
|
| 436 |
+
image = pipe(prompt).images[0]
|
| 437 |
+
```
|
| 438 |
+
You can use our distilled model to generate images even faster:
|
| 439 |
+
|
| 440 |
+
```py
|
| 441 |
+
import torch
|
| 442 |
+
from diffusers import HunyuanDiTPipeline
|
| 443 |
+
|
| 444 |
+
pipe = HunyuanDiTPipeline.from_pretrained("Tencent-Hunyuan/HunyuanDiT-Diffusers-Distilled", torch_dtype=torch.float16)
|
| 445 |
+
pipe.to("cuda")
|
| 446 |
+
|
| 447 |
+
# You may also use English prompt as HunyuanDiT supports both English and Chinese
|
| 448 |
+
# prompt = "An astronaut riding a horse"
|
| 449 |
+
prompt = "ไธไธชๅฎ่ชๅๅจ้ช้ฉฌ"
|
| 450 |
+
image = pipe(prompt, num_inference_steps=25).images[0]
|
| 451 |
+
```
|
| 452 |
+
More details can be found in [HunyuanDiT-Diffusers-Distilled](https://huggingface.co/Tencent-Hunyuan/HunyuanDiT-Diffusers-Distilled)
|
| 453 |
+
|
| 454 |
+
### Using Command Line
|
| 455 |
+
|
| 456 |
+
We provide several commands to quick start:
|
| 457 |
+
|
| 458 |
+
```shell
|
| 459 |
+
# Prompt Enhancement + Text-to-Image. Torch mode
|
| 460 |
+
python sample_t2i.py --prompt "ๆธ่ๅฑๆ"
|
| 461 |
+
|
| 462 |
+
# Only Text-to-Image. Torch mode
|
| 463 |
+
python sample_t2i.py --prompt "ๆธ่ๅฑๆ" --no-enhance
|
| 464 |
+
|
| 465 |
+
# Only Text-to-Image. Flash Attention mode
|
| 466 |
+
python sample_t2i.py --infer-mode fa --prompt "ๆธ่ๅฑๆ"
|
| 467 |
+
|
| 468 |
+
# Generate an image with other image sizes.
|
| 469 |
+
python sample_t2i.py --prompt "ๆธ่ๅฑๆ" --image-size 1280 768
|
| 470 |
+
|
| 471 |
+
# Prompt Enhancement + Text-to-Image. DialogGen loads with 4-bit quantization, but it may loss performance.
|
| 472 |
+
python sample_t2i.py --prompt "ๆธ่ๅฑๆ" --load-4bit
|
| 473 |
+
|
| 474 |
+
```
|
| 475 |
+
|
| 476 |
+
More example prompts can be found in [example_prompts.txt](example_prompts.txt)
|
| 477 |
+
|
| 478 |
+
### More Configurations
|
| 479 |
+
|
| 480 |
+
We list some more useful configurations for easy usage:
|
| 481 |
+
|
| 482 |
+
| Argument | Default | Description |
|
| 483 |
+
|:---------------:|:---------:|:---------------------------------------------------:|
|
| 484 |
+
| `--prompt` | None | The text prompt for image generation |
|
| 485 |
+
| `--image-size` | 1024 1024 | The size of the generated image |
|
| 486 |
+
| `--seed` | 42 | The random seed for generating images |
|
| 487 |
+
| `--infer-steps` | 100 | The number of steps for sampling |
|
| 488 |
+
| `--negative` | - | The negative prompt for image generation |
|
| 489 |
+
| `--infer-mode` | torch | The inference mode (torch, fa, or trt) |
|
| 490 |
+
| `--sampler` | ddpm | The diffusion sampler (ddpm, ddim, or dpmms) |
|
| 491 |
+
| `--no-enhance` | False | Disable the prompt enhancement model |
|
| 492 |
+
| `--model-root` | ckpts | The root directory of the model checkpoints |
|
| 493 |
+
| `--load-key` | ema | Load the student model or EMA model (ema or module) |
|
| 494 |
+
| `--load-4bit` | Fasle | Load DialogGen model with 4bit quantization |
|
| 495 |
+
|
| 496 |
+
### Using ComfyUI
|
| 497 |
+
|
| 498 |
+
We provide several commands to quick start:
|
| 499 |
+
|
| 500 |
+
```shell
|
| 501 |
+
# Download comfyui code
|
| 502 |
+
git clone https://github.com/comfyanonymous/ComfyUI.git
|
| 503 |
+
|
| 504 |
+
# Install torch, torchvision, torchaudio
|
| 505 |
+
pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cu117
|
| 506 |
+
|
| 507 |
+
# Install Comfyui essential python package
|
| 508 |
+
cd ComfyUI
|
| 509 |
+
pip install -r requirements.txt
|
| 510 |
+
|
| 511 |
+
# ComfyUI has been successfully installed!
|
| 512 |
+
|
| 513 |
+
# Download model weight as before or link the existing model folder to ComfyUI.
|
| 514 |
+
python -m pip install "huggingface_hub[cli]"
|
| 515 |
+
mkdir models/hunyuan
|
| 516 |
+
huggingface-cli download Tencent-Hunyuan/HunyuanDiT --local-dir ./models/hunyuan/ckpts
|
| 517 |
+
|
| 518 |
+
# Move to the ComfyUI custom_nodes folder and copy comfyui-hydit folder from HunyuanDiT Repo.
|
| 519 |
+
cd custom_nodes
|
| 520 |
+
cp -r ${HunyuanDiT}/comfyui-hydit ./
|
| 521 |
+
cd comfyui-hydit
|
| 522 |
+
|
| 523 |
+
# Install some essential python Package.
|
| 524 |
+
pip install -r requirements.txt
|
| 525 |
+
|
| 526 |
+
# Our tool has been successfully installed!
|
| 527 |
+
|
| 528 |
+
# Go to ComfyUI main folder
|
| 529 |
+
cd ../..
|
| 530 |
+
# Run the ComfyUI Lauch command
|
| 531 |
+
python main.py --listen --port 80
|
| 532 |
+
|
| 533 |
+
# Running ComfyUI successfully!
|
| 534 |
+
```
|
| 535 |
+
More details can be found in [ComfyUI README](comfyui-hydit/README.md)
|
| 536 |
+
|
| 537 |
+
## ๐ Acceleration (for Linux)
|
| 538 |
+
|
| 539 |
+
- We provide TensorRT version of HunyuanDiT for inference acceleration (faster than flash attention).
|
| 540 |
+
See [Tencent-Hunyuan/TensorRT-libs](https://huggingface.co/Tencent-Hunyuan/TensorRT-libs) for more details.
|
| 541 |
+
|
| 542 |
+
- We provide Distillation version of HunyuanDiT for inference acceleration.
|
| 543 |
+
See [Tencent-Hunyuan/Distillation](https://huggingface.co/Tencent-Hunyuan/Distillation) for more details.
|
| 544 |
+
|
| 545 |
+
## ๐ BibTeX
|
| 546 |
+
If you find [Hunyuan-DiT](https://arxiv.org/abs/2405.08748) or [DialogGen](https://arxiv.org/abs/2403.08857) useful for your research and applications, please cite using this BibTeX:
|
| 547 |
+
|
| 548 |
+
```BibTeX
|
| 549 |
+
@misc{li2024hunyuandit,
|
| 550 |
+
title={Hunyuan-DiT: A Powerful Multi-Resolution Diffusion Transformer with Fine-Grained Chinese Understanding},
|
| 551 |
+
author={Zhimin Li and Jianwei Zhang and Qin Lin and Jiangfeng Xiong and Yanxin Long and Xinchi Deng and Yingfang Zhang and Xingchao Liu and Minbin Huang and Zedong Xiao and Dayou Chen and Jiajun He and Jiahao Li and Wenyue Li and Chen Zhang and Rongwei Quan and Jianxiang Lu and Jiabin Huang and Xiaoyan Yuan and Xiaoxiao Zheng and Yixuan Li and Jihong Zhang and Chao Zhang and Meng Chen and Jie Liu and Zheng Fang and Weiyan Wang and Jinbao Xue and Yangyu Tao and Jianchen Zhu and Kai Liu and Sihuan Lin and Yifu Sun and Yun Li and Dongdong Wang and Mingtao Chen and Zhichao Hu and Xiao Xiao and Yan Chen and Yuhong Liu and Wei Liu and Di Wang and Yong Yang and Jie Jiang and Qinglin Lu},
|
| 552 |
+
year={2024},
|
| 553 |
+
eprint={2405.08748},
|
| 554 |
+
archivePrefix={arXiv},
|
| 555 |
+
primaryClass={cs.CV}
|
| 556 |
+
}
|
| 557 |
+
|
| 558 |
+
@article{huang2024dialoggen,
|
| 559 |
+
title={DialogGen: Multi-modal Interactive Dialogue System for Multi-turn Text-to-Image Generation},
|
| 560 |
+
author={Huang, Minbin and Long, Yanxin and Deng, Xinchi and Chu, Ruihang and Xiong, Jiangfeng and Liang, Xiaodan and Cheng, Hong and Lu, Qinglin and Liu, Wei},
|
| 561 |
+
journal={arXiv preprint arXiv:2403.08857},
|
| 562 |
+
year={2024}
|
| 563 |
+
}
|
| 564 |
+
```
|
| 565 |
+
|
| 566 |
+
## Start History
|
| 567 |
+
|
| 568 |
+
<a href="https://star-history.com/#Tencent/HunyuanDiT&Date">
|
| 569 |
+
<picture>
|
| 570 |
+
<source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/svg?repos=Tencent/HunyuanDiT&type=Date&theme=dark" />
|
| 571 |
+
<source media="(prefers-color-scheme: light)" srcset="https://api.star-history.com/svg?repos=Tencent/HunyuanDiT&type=Date" />
|
| 572 |
+
<img alt="Star History Chart" src="https://api.star-history.com/svg?repos=Tencent/HunyuanDiT&type=Date" />
|
| 573 |
+
</picture>
|
| 574 |
+
</a>
|
|
|
|
|
|
|
|
|
|
|
|