Spaces:
Paused
Paused
File size: 22,540 Bytes
186aa49 | 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 29 30 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 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 | # Copyright 2026 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from ...utils import logging
from ..modular_pipeline import AutoPipelineBlocks, SequentialPipelineBlocks
from ..modular_pipeline_utils import InsertableDict, OutputParam
from .before_denoise import (
FluxImg2ImgPrepareLatentsStep,
FluxImg2ImgSetTimestepsStep,
FluxPrepareLatentsStep,
FluxRoPEInputsStep,
FluxSetTimestepsStep,
)
from .decoders import FluxDecodeStep
from .denoise import FluxDenoiseStep
from .encoders import (
FluxProcessImagesInputStep,
FluxTextEncoderStep,
FluxVaeEncoderStep,
)
from .inputs import (
FluxAdditionalInputsStep,
FluxTextInputStep,
)
logger = logging.get_logger(__name__) # pylint: disable=invalid-name
# vae encoder (run before before_denoise)
# auto_docstring
class FluxImg2ImgVaeEncoderStep(SequentialPipelineBlocks):
"""
Vae encoder step that preprocess andencode the image inputs into their latent representations.
Components:
image_processor (`VaeImageProcessor`) vae (`AutoencoderKL`)
Inputs:
resized_image (`None`, *optional*):
TODO: Add description.
image (`None`, *optional*):
TODO: Add description.
height (`None`, *optional*):
TODO: Add description.
width (`None`, *optional*):
TODO: Add description.
generator (`None`, *optional*):
TODO: Add description.
Outputs:
processed_image (`None`):
TODO: Add description.
image_latents (`Tensor`):
The latents representing the reference image
"""
model_name = "flux"
block_classes = [FluxProcessImagesInputStep(), FluxVaeEncoderStep()]
block_names = ["preprocess", "encode"]
@property
def description(self) -> str:
return "Vae encoder step that preprocess andencode the image inputs into their latent representations."
# auto_docstring
class FluxAutoVaeEncoderStep(AutoPipelineBlocks):
"""
Vae encoder step that encode the image inputs into their latent representations.
This is an auto pipeline block that works for img2img tasks.
- `FluxImg2ImgVaeEncoderStep` (img2img) is used when only `image` is provided. - if `image` is not provided,
step will be skipped.
Components:
image_processor (`VaeImageProcessor`) vae (`AutoencoderKL`)
Inputs:
resized_image (`None`, *optional*):
TODO: Add description.
image (`None`, *optional*):
TODO: Add description.
height (`None`, *optional*):
TODO: Add description.
width (`None`, *optional*):
TODO: Add description.
generator (`None`, *optional*):
TODO: Add description.
Outputs:
processed_image (`None`):
TODO: Add description.
image_latents (`Tensor`):
The latents representing the reference image
"""
model_name = "flux"
block_classes = [FluxImg2ImgVaeEncoderStep]
block_names = ["img2img"]
block_trigger_inputs = ["image"]
@property
def description(self):
return (
"Vae encoder step that encode the image inputs into their latent representations.\n"
+ "This is an auto pipeline block that works for img2img tasks.\n"
+ " - `FluxImg2ImgVaeEncoderStep` (img2img) is used when only `image` is provided."
+ " - if `image` is not provided, step will be skipped."
)
# before_denoise: text2img
# auto_docstring
class FluxBeforeDenoiseStep(SequentialPipelineBlocks):
"""
Before denoise step that prepares the inputs for the denoise step in text-to-image generation.
Components:
scheduler (`FlowMatchEulerDiscreteScheduler`)
Inputs:
height (`int`, *optional*):
TODO: Add description.
width (`int`, *optional*):
TODO: Add description.
latents (`Tensor | NoneType`, *optional*):
TODO: Add description.
num_images_per_prompt (`int`, *optional*, defaults to 1):
TODO: Add description.
generator (`None`, *optional*):
TODO: Add description.
batch_size (`int`):
Number of prompts, the final batch size of model inputs should be `batch_size * num_images_per_prompt`.
Can be generated in input step.
dtype (`dtype`, *optional*):
The dtype of the model inputs
num_inference_steps (`None`, *optional*, defaults to 50):
TODO: Add description.
timesteps (`None`, *optional*):
TODO: Add description.
sigmas (`None`, *optional*):
TODO: Add description.
guidance_scale (`None`, *optional*, defaults to 3.5):
TODO: Add description.
prompt_embeds (`None`, *optional*):
TODO: Add description.
Outputs:
latents (`Tensor`):
The initial latents to use for the denoising process
timesteps (`Tensor`):
The timesteps to use for inference
num_inference_steps (`int`):
The number of denoising steps to perform at inference time
guidance (`Tensor`):
Optional guidance to be used.
txt_ids (`list`):
The sequence lengths of the prompt embeds, used for RoPE calculation.
img_ids (`list`):
The sequence lengths of the image latents, used for RoPE calculation.
"""
model_name = "flux"
block_classes = [FluxPrepareLatentsStep(), FluxSetTimestepsStep(), FluxRoPEInputsStep()]
block_names = ["prepare_latents", "set_timesteps", "prepare_rope_inputs"]
@property
def description(self):
return "Before denoise step that prepares the inputs for the denoise step in text-to-image generation."
# before_denoise: img2img
# auto_docstring
class FluxImg2ImgBeforeDenoiseStep(SequentialPipelineBlocks):
"""
Before denoise step that prepare the inputs for the denoise step for img2img task.
Components:
scheduler (`FlowMatchEulerDiscreteScheduler`)
Inputs:
height (`int`, *optional*):
TODO: Add description.
width (`int`, *optional*):
TODO: Add description.
latents (`Tensor | NoneType`, *optional*):
TODO: Add description.
num_images_per_prompt (`int`, *optional*, defaults to 1):
TODO: Add description.
generator (`None`, *optional*):
TODO: Add description.
batch_size (`int`):
Number of prompts, the final batch size of model inputs should be `batch_size * num_images_per_prompt`.
Can be generated in input step.
dtype (`dtype`, *optional*):
The dtype of the model inputs
num_inference_steps (`None`, *optional*, defaults to 50):
TODO: Add description.
timesteps (`None`, *optional*):
TODO: Add description.
sigmas (`None`, *optional*):
TODO: Add description.
strength (`None`, *optional*, defaults to 0.6):
TODO: Add description.
guidance_scale (`None`, *optional*, defaults to 3.5):
TODO: Add description.
image_latents (`Tensor`):
The image latents to use for the denoising process. Can be generated in vae encoder and packed in input
step.
prompt_embeds (`None`, *optional*):
TODO: Add description.
Outputs:
latents (`Tensor`):
The initial latents to use for the denoising process
timesteps (`Tensor`):
The timesteps to use for inference
num_inference_steps (`int`):
The number of denoising steps to perform at inference time
guidance (`Tensor`):
Optional guidance to be used.
initial_noise (`Tensor`):
The initial random noised used for inpainting denoising.
txt_ids (`list`):
The sequence lengths of the prompt embeds, used for RoPE calculation.
img_ids (`list`):
The sequence lengths of the image latents, used for RoPE calculation.
"""
model_name = "flux"
block_classes = [
FluxPrepareLatentsStep(),
FluxImg2ImgSetTimestepsStep(),
FluxImg2ImgPrepareLatentsStep(),
FluxRoPEInputsStep(),
]
block_names = ["prepare_latents", "set_timesteps", "prepare_img2img_latents", "prepare_rope_inputs"]
@property
def description(self):
return "Before denoise step that prepare the inputs for the denoise step for img2img task."
# before_denoise: all task (text2img, img2img)
# auto_docstring
class FluxAutoBeforeDenoiseStep(AutoPipelineBlocks):
"""
Before denoise step that prepare the inputs for the denoise step.
This is an auto pipeline block that works for text2image.
- `FluxBeforeDenoiseStep` (text2image) is used.
- `FluxImg2ImgBeforeDenoiseStep` (img2img) is used when only `image_latents` is provided.
Components:
scheduler (`FlowMatchEulerDiscreteScheduler`)
Inputs:
height (`int`):
TODO: Add description.
width (`int`):
TODO: Add description.
latents (`Tensor | NoneType`, *optional*):
TODO: Add description.
num_images_per_prompt (`int`, *optional*, defaults to 1):
TODO: Add description.
generator (`None`, *optional*):
TODO: Add description.
batch_size (`int`):
Number of prompts, the final batch size of model inputs should be `batch_size * num_images_per_prompt`.
Can be generated in input step.
dtype (`dtype`, *optional*):
The dtype of the model inputs
num_inference_steps (`None`, *optional*, defaults to 50):
TODO: Add description.
timesteps (`None`, *optional*):
TODO: Add description.
sigmas (`None`, *optional*):
TODO: Add description.
strength (`None`, *optional*, defaults to 0.6):
TODO: Add description.
guidance_scale (`None`, *optional*, defaults to 3.5):
TODO: Add description.
image_latents (`Tensor`, *optional*):
The image latents to use for the denoising process. Can be generated in vae encoder and packed in input
step.
prompt_embeds (`None`, *optional*):
TODO: Add description.
Outputs:
latents (`Tensor`):
The initial latents to use for the denoising process
timesteps (`Tensor`):
The timesteps to use for inference
num_inference_steps (`int`):
The number of denoising steps to perform at inference time
guidance (`Tensor`):
Optional guidance to be used.
initial_noise (`Tensor`):
The initial random noised used for inpainting denoising.
txt_ids (`list`):
The sequence lengths of the prompt embeds, used for RoPE calculation.
img_ids (`list`):
The sequence lengths of the image latents, used for RoPE calculation.
"""
model_name = "flux"
block_classes = [FluxImg2ImgBeforeDenoiseStep, FluxBeforeDenoiseStep]
block_names = ["img2img", "text2image"]
block_trigger_inputs = ["image_latents", None]
@property
def description(self):
return (
"Before denoise step that prepare the inputs for the denoise step.\n"
+ "This is an auto pipeline block that works for text2image.\n"
+ " - `FluxBeforeDenoiseStep` (text2image) is used.\n"
+ " - `FluxImg2ImgBeforeDenoiseStep` (img2img) is used when only `image_latents` is provided.\n"
)
# inputs: text2image/img2img
# auto_docstring
class FluxImg2ImgInputStep(SequentialPipelineBlocks):
"""
Input step that prepares the inputs for the img2img denoising step. It:
Inputs:
num_images_per_prompt (`None`, *optional*, defaults to 1):
TODO: Add description.
prompt_embeds (`Tensor`):
Pre-generated text embeddings. Can be generated from text_encoder step.
pooled_prompt_embeds (`Tensor`, *optional*):
Pre-generated pooled text embeddings. Can be generated from text_encoder step.
height (`None`, *optional*):
TODO: Add description.
width (`None`, *optional*):
TODO: Add description.
image_latents (`None`, *optional*):
TODO: Add description.
Outputs:
batch_size (`int`):
Number of prompts, the final batch size of model inputs should be batch_size * num_images_per_prompt
dtype (`dtype`):
Data type of model tensor inputs (determined by `prompt_embeds`)
prompt_embeds (`Tensor`):
text embeddings used to guide the image generation
pooled_prompt_embeds (`Tensor`):
pooled text embeddings used to guide the image generation
image_height (`int`):
The height of the image latents
image_width (`int`):
The width of the image latents
"""
model_name = "flux"
block_classes = [FluxTextInputStep(), FluxAdditionalInputsStep()]
block_names = ["text_inputs", "additional_inputs"]
@property
def description(self):
return "Input step that prepares the inputs for the img2img denoising step. It:\n"
" - make sure the text embeddings have consistent batch size as well as the additional inputs (`image_latents`).\n"
" - update height/width based `image_latents`, patchify `image_latents`."
# auto_docstring
class FluxAutoInputStep(AutoPipelineBlocks):
"""
Input step that standardize the inputs for the denoising step, e.g. make sure inputs have consistent batch size,
and patchified.
This is an auto pipeline block that works for text2image/img2img tasks.
- `FluxImg2ImgInputStep` (img2img) is used when `image_latents` is provided.
- `FluxTextInputStep` (text2image) is used when `image_latents` are not provided.
Inputs:
num_images_per_prompt (`None`, *optional*, defaults to 1):
TODO: Add description.
prompt_embeds (`Tensor`):
Pre-generated text embeddings. Can be generated from text_encoder step.
pooled_prompt_embeds (`Tensor`, *optional*):
Pre-generated pooled text embeddings. Can be generated from text_encoder step.
height (`None`, *optional*):
TODO: Add description.
width (`None`, *optional*):
TODO: Add description.
image_latents (`None`, *optional*):
TODO: Add description.
Outputs:
batch_size (`int`):
Number of prompts, the final batch size of model inputs should be batch_size * num_images_per_prompt
dtype (`dtype`):
Data type of model tensor inputs (determined by `prompt_embeds`)
prompt_embeds (`Tensor`):
text embeddings used to guide the image generation
pooled_prompt_embeds (`Tensor`):
pooled text embeddings used to guide the image generation
image_height (`int`):
The height of the image latents
image_width (`int`):
The width of the image latents
"""
model_name = "flux"
block_classes = [FluxImg2ImgInputStep, FluxTextInputStep]
block_names = ["img2img", "text2image"]
block_trigger_inputs = ["image_latents", None]
@property
def description(self):
return (
"Input step that standardize the inputs for the denoising step, e.g. make sure inputs have consistent batch size, and patchified. \n"
" This is an auto pipeline block that works for text2image/img2img tasks.\n"
+ " - `FluxImg2ImgInputStep` (img2img) is used when `image_latents` is provided.\n"
+ " - `FluxTextInputStep` (text2image) is used when `image_latents` are not provided.\n"
)
# auto_docstring
class FluxCoreDenoiseStep(SequentialPipelineBlocks):
"""
Core step that performs the denoising process for Flux.
This step supports text-to-image and image-to-image tasks for Flux:
- for image-to-image generation, you need to provide `image_latents`
- for text-to-image generation, all you need to provide is prompt embeddings.
Components:
scheduler (`FlowMatchEulerDiscreteScheduler`) transformer (`FluxTransformer2DModel`)
Inputs:
num_images_per_prompt (`None`, *optional*, defaults to 1):
TODO: Add description.
prompt_embeds (`Tensor`):
Pre-generated text embeddings. Can be generated from text_encoder step.
pooled_prompt_embeds (`Tensor`, *optional*):
Pre-generated pooled text embeddings. Can be generated from text_encoder step.
height (`None`, *optional*):
TODO: Add description.
width (`None`, *optional*):
TODO: Add description.
image_latents (`None`, *optional*):
TODO: Add description.
latents (`Tensor | NoneType`, *optional*):
TODO: Add description.
generator (`None`, *optional*):
TODO: Add description.
num_inference_steps (`None`, *optional*, defaults to 50):
TODO: Add description.
timesteps (`None`, *optional*):
TODO: Add description.
sigmas (`None`, *optional*):
TODO: Add description.
strength (`None`, *optional*, defaults to 0.6):
TODO: Add description.
guidance_scale (`None`, *optional*, defaults to 3.5):
TODO: Add description.
joint_attention_kwargs (`None`, *optional*):
TODO: Add description.
Outputs:
latents (`Tensor`):
Denoised latents.
"""
model_name = "flux"
block_classes = [FluxAutoInputStep, FluxAutoBeforeDenoiseStep, FluxDenoiseStep]
block_names = ["input", "before_denoise", "denoise"]
@property
def description(self):
return (
"Core step that performs the denoising process for Flux.\n"
+ "This step supports text-to-image and image-to-image tasks for Flux:\n"
+ " - for image-to-image generation, you need to provide `image_latents`\n"
+ " - for text-to-image generation, all you need to provide is prompt embeddings."
)
@property
def outputs(self):
return [
OutputParam.template("latents"),
]
# Auto blocks (text2image and img2img)
AUTO_BLOCKS = InsertableDict(
[
("text_encoder", FluxTextEncoderStep()),
("vae_encoder", FluxAutoVaeEncoderStep()),
("denoise", FluxCoreDenoiseStep()),
("decode", FluxDecodeStep()),
]
)
# auto_docstring
class FluxAutoBlocks(SequentialPipelineBlocks):
"""
Auto Modular pipeline for text-to-image and image-to-image using Flux.
Supported workflows:
- `text2image`: requires `prompt`
- `image2image`: requires `image`, `prompt`
Components:
text_encoder (`CLIPTextModel`) tokenizer (`CLIPTokenizer`) text_encoder_2 (`T5EncoderModel`) tokenizer_2
(`T5Tokenizer`) image_processor (`VaeImageProcessor`) vae (`AutoencoderKL`) scheduler
(`FlowMatchEulerDiscreteScheduler`) transformer (`FluxTransformer2DModel`)
Inputs:
prompt (`None`, *optional*):
TODO: Add description.
prompt_2 (`None`, *optional*):
TODO: Add description.
max_sequence_length (`int`, *optional*, defaults to 512):
TODO: Add description.
joint_attention_kwargs (`None`, *optional*):
TODO: Add description.
resized_image (`None`, *optional*):
TODO: Add description.
image (`None`, *optional*):
TODO: Add description.
height (`None`, *optional*):
TODO: Add description.
width (`None`, *optional*):
TODO: Add description.
generator (`None`, *optional*):
TODO: Add description.
num_images_per_prompt (`None`, *optional*, defaults to 1):
TODO: Add description.
image_latents (`None`, *optional*):
TODO: Add description.
latents (`Tensor | NoneType`, *optional*):
TODO: Add description.
num_inference_steps (`None`, *optional*, defaults to 50):
TODO: Add description.
timesteps (`None`, *optional*):
TODO: Add description.
sigmas (`None`, *optional*):
TODO: Add description.
strength (`None`, *optional*, defaults to 0.6):
TODO: Add description.
guidance_scale (`None`, *optional*, defaults to 3.5):
TODO: Add description.
output_type (`None`, *optional*, defaults to pil):
TODO: Add description.
Outputs:
images (`list`):
Generated images.
"""
model_name = "flux"
block_classes = AUTO_BLOCKS.values()
block_names = AUTO_BLOCKS.keys()
_workflow_map = {
"text2image": {"prompt": True},
"image2image": {"image": True, "prompt": True},
}
@property
def description(self):
return "Auto Modular pipeline for text-to-image and image-to-image using Flux."
@property
def outputs(self):
return [OutputParam.template("images")]
|