File size: 17,357 Bytes
ac2243f | 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 | <!--Copyright 2025 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.
-->
# Building Custom Blocks
[ModularPipelineBlocks](./pipeline_block) are the fundamental building blocks of a [`ModularPipeline`]. You can create custom blocks by defining their inputs, outputs, and computation logic. This guide demonstrates how to create and use a custom block.
> [!TIP]
> Explore the [Modular Diffusers Custom Blocks](https://huggingface.co/collections/diffusers/modular-diffusers-custom-blocks) collection for official custom modular blocks like Nano Banana.
## Project Structure
Your custom block project should use the following structure:
```shell
.
├── block.py
└── modular_config.json
```
- `block.py` contains the custom block implementation
- `modular_config.json` contains the metadata needed to load the block
## Example: Florence 2 Inpainting Block
In this example we will create a custom block that uses the [Florence 2](https://huggingface.co/docs/transformers/model_doc/florence2) model to process an input image and generate a mask for inpainting.
The first step is to define the components that the block will use. In this case, we will need to use the `Florence2ForConditionalGeneration` model and its corresponding processor `AutoProcessor`. When defining components, we must specify the name of the component within our pipeline, model class via `type_hint`, and provide a `pretrained_model_name_or_path` for the component if we intend to load the model weights from a specific repository on the Hub.
```py
# Inside block.py
from diffusers.modular_pipelines import (
ModularPipelineBlocks,
ComponentSpec,
)
from transformers import AutoProcessor, Florence2ForConditionalGeneration
class Florence2ImageAnnotatorBlock(ModularPipelineBlocks):
@property
def expected_components(self):
return [
ComponentSpec(
name="image_annotator",
type_hint=Florence2ForConditionalGeneration,
pretrained_model_name_or_path="florence-community/Florence-2-base-ft",
),
ComponentSpec(
name="image_annotator_processor",
type_hint=AutoProcessor,
pretrained_model_name_or_path="florence-community/Florence-2-base-ft",
),
]
```
Next, we define the inputs and outputs of the block. The inputs include the image to be annotated, the annotation task, and the annotation prompt. The outputs include the generated mask image and annotations.
```py
from typing import List, Union
from PIL import Image, ImageDraw
import torch
import numpy as np
from diffusers.modular_pipelines import (
PipelineState,
ModularPipelineBlocks,
InputParam,
ComponentSpec,
OutputParam,
)
from transformers import AutoProcessor, Florence2ForConditionalGeneration
class Florence2ImageAnnotatorBlock(ModularPipelineBlocks):
@property
def expected_components(self):
return [
ComponentSpec(
name="image_annotator",
type_hint=Florence2ForConditionalGeneration,
pretrained_model_name_or_path="florence-community/Florence-2-base-ft",
),
ComponentSpec(
name="image_annotator_processor",
type_hint=AutoProcessor,
pretrained_model_name_or_path="florence-community/Florence-2-base-ft",
),
]
@property
def inputs(self) -> List[InputParam]:
return [
InputParam(
"image",
type_hint=Union[Image.Image, List[Image.Image]],
required=True,
description="Image(s) to annotate",
),
InputParam(
"annotation_task",
type_hint=Union[str, List[str]],
required=True,
default="<REFERRING_EXPRESSION_SEGMENTATION>",
description="""Annotation Task to perform on the image.
Supported Tasks:
<OD>
<REFERRING_EXPRESSION_SEGMENTATION>
<CAPTION>
<DETAILED_CAPTION>
<MORE_DETAILED_CAPTION>
<DENSE_REGION_CAPTION>
<CAPTION_TO_PHRASE_GROUNDING>
<OPEN_VOCABULARY_DETECTION>
""",
),
InputParam(
"annotation_prompt",
type_hint=Union[str, List[str]],
required=True,
description="""Annotation Prompt to provide more context to the task.
Can be used to detect or segment out specific elements in the image
""",
),
InputParam(
"annotation_output_type",
type_hint=str,
required=True,
default="mask_image",
description="""Output type from annotation predictions. Available options are
mask_image:
-black and white mask image for the given image based on the task type
mask_overlay:
- mask overlayed on the original image
bounding_box:
- bounding boxes drawn on the original image
""",
),
InputParam(
"annotation_overlay",
type_hint=bool,
required=True,
default=False,
description="",
),
]
@property
def intermediate_outputs(self) -> List[OutputParam]:
return [
OutputParam(
"mask_image",
type_hint=Image,
description="Inpainting Mask for input Image(s)",
),
OutputParam(
"annotations",
type_hint=dict,
description="Annotations Predictions for input Image(s)",
),
OutputParam(
"image",
type_hint=Image,
description="Annotated input Image(s)",
),
]
```
Now we implement the `__call__` method, which contains the logic for processing the input image and generating the mask.
```py
from typing import List, Union
from PIL import Image, ImageDraw
import torch
import numpy as np
from diffusers.modular_pipelines import (
PipelineState,
ModularPipelineBlocks,
InputParam,
ComponentSpec,
OutputParam,
)
from transformers import AutoProcessor, Florence2ForConditionalGeneration
class Florence2ImageAnnotatorBlock(ModularPipelineBlocks):
@property
def expected_components(self):
return [
ComponentSpec(
name="image_annotator",
type_hint=Florence2ForConditionalGeneration,
pretrained_model_name_or_path="florence-community/Florence-2-base-ft",
),
ComponentSpec(
name="image_annotator_processor",
type_hint=AutoProcessor,
pretrained_model_name_or_path="florence-community/Florence-2-base-ft",
),
]
@property
def inputs(self) -> List[InputParam]:
return [
InputParam(
"image",
type_hint=Union[Image.Image, List[Image.Image]],
required=True,
description="Image(s) to annotate",
),
InputParam(
"annotation_task",
type_hint=Union[str, List[str]],
required=True,
default="<REFERRING_EXPRESSION_SEGMENTATION>",
description="""Annotation Task to perform on the image.
Supported Tasks:
<OD>
<REFERRING_EXPRESSION_SEGMENTATION>
<CAPTION>
<DETAILED_CAPTION>
<MORE_DETAILED_CAPTION>
<DENSE_REGION_CAPTION>
<CAPTION_TO_PHRASE_GROUNDING>
<OPEN_VOCABULARY_DETECTION>
""",
),
InputParam(
"annotation_prompt",
type_hint=Union[str, List[str]],
required=True,
description="""Annotation Prompt to provide more context to the task.
Can be used to detect or segment out specific elements in the image
""",
),
InputParam(
"annotation_output_type",
type_hint=str,
required=True,
default="mask_image",
description="""Output type from annotation predictions. Available options are
mask_image:
-black and white mask image for the given image based on the task type
mask_overlay:
- mask overlayed on the original image
bounding_box:
- bounding boxes drawn on the original image
""",
),
InputParam(
"annotation_overlay",
type_hint=bool,
required=True,
default=False,
description="",
),
]
@property
def intermediate_outputs(self) -> List[OutputParam]:
return [
OutputParam(
"mask_image",
type_hint=Image,
description="Inpainting Mask for input Image(s)",
),
OutputParam(
"annotations",
type_hint=dict,
description="Annotations Predictions for input Image(s)",
),
OutputParam(
"image",
type_hint=Image,
description="Annotated input Image(s)",
),
]
def get_annotations(self, components, images, prompts, task):
task_prompts = [task + prompt for prompt in prompts]
inputs = components.image_annotator_processor(
text=task_prompts, images=images, return_tensors="pt"
).to(components.image_annotator.device, components.image_annotator.dtype)
generated_ids = components.image_annotator.generate(
input_ids=inputs["input_ids"],
pixel_values=inputs["pixel_values"],
max_new_tokens=1024,
early_stopping=False,
do_sample=False,
num_beams=3,
)
annotations = components.image_annotator_processor.batch_decode(
generated_ids, skip_special_tokens=False
)
outputs = []
for image, annotation in zip(images, annotations):
outputs.append(
components.image_annotator_processor.post_process_generation(
annotation, task=task, image_size=(image.width, image.height)
)
)
return outputs
def prepare_mask(self, images, annotations, overlay=False, fill="white"):
masks = []
for image, annotation in zip(images, annotations):
mask_image = image.copy() if overlay else Image.new("L", image.size, 0)
draw = ImageDraw.Draw(mask_image)
for _, _annotation in annotation.items():
if "polygons" in _annotation:
for polygon in _annotation["polygons"]:
polygon = np.array(polygon).reshape(-1, 2)
if len(polygon) < 3:
continue
polygon = polygon.reshape(-1).tolist()
draw.polygon(polygon, fill=fill)
elif "bbox" in _annotation:
bbox = _annotation["bbox"]
draw.rectangle(bbox, fill="white")
masks.append(mask_image)
return masks
def prepare_bounding_boxes(self, images, annotations):
outputs = []
for image, annotation in zip(images, annotations):
image_copy = image.copy()
draw = ImageDraw.Draw(image_copy)
for _, _annotation in annotation.items():
bbox = _annotation["bbox"]
label = _annotation["label"]
draw.rectangle(bbox, outline="red", width=3)
draw.text((bbox[0], bbox[1] - 20), label, fill="red")
outputs.append(image_copy)
return outputs
def prepare_inputs(self, images, prompts):
prompts = prompts or ""
if isinstance(images, Image.Image):
images = [images]
if isinstance(prompts, str):
prompts = [prompts]
if len(images) != len(prompts):
raise ValueError("Number of images and annotation prompts must match.")
return images, prompts
@torch.no_grad()
def __call__(self, components, state: PipelineState) -> PipelineState:
block_state = self.get_block_state(state)
images, annotation_task_prompt = self.prepare_inputs(
block_state.image, block_state.annotation_prompt
)
task = block_state.annotation_task
fill = block_state.fill
annotations = self.get_annotations(
components, images, annotation_task_prompt, task
)
block_state.annotations = annotations
if block_state.annotation_output_type == "mask_image":
block_state.mask_image = self.prepare_mask(images, annotations)
else:
block_state.mask_image = None
if block_state.annotation_output_type == "mask_overlay":
block_state.image = self.prepare_mask(images, annotations, overlay=True, fill=fill)
elif block_state.annotation_output_type == "bounding_box":
block_state.image = self.prepare_bounding_boxes(images, annotations)
self.set_block_state(state, block_state)
return components, state
```
Once we have defined our custom block, we can save it to the Hub, using either the CLI or the [`push_to_hub`] method. This will make it easy to share and reuse our custom block with other pipelines.
<hfoptions id="share">
<hfoption id="hf CLI">
```shell
# In the folder with the `block.py` file, run:
diffusers-cli custom_block
```
Then upload the block to the Hub:
```shell
hf upload <your repo id> . .
```
</hfoption>
<hfoption id="push_to_hub">
```py
from block import Florence2ImageAnnotatorBlock
block = Florence2ImageAnnotatorBlock()
block.push_to_hub("<your repo id>")
```
</hfoption>
</hfoptions>
## Using Custom Blocks
Load the custom block with [`~ModularPipelineBlocks.from_pretrained`] and set `trust_remote_code=True`.
```py
import torch
from diffusers.modular_pipelines import ModularPipelineBlocks, SequentialPipelineBlocks
from diffusers.modular_pipelines.stable_diffusion_xl import INPAINT_BLOCKS
from diffusers.utils import load_image
# Fetch the Florence2 image annotator block that will create our mask
image_annotator_block = ModularPipelineBlocks.from_pretrained("diffusers/florence-2-custom-block", trust_remote_code=True)
my_blocks = INPAINT_BLOCKS.copy()
# insert the annotation block before the image encoding step
my_blocks.insert("image_annotator", image_annotator_block, 1)
# Create our initial set of inpainting blocks
blocks = SequentialPipelineBlocks.from_blocks_dict(my_blocks)
repo_id = "diffusers/modular-stable-diffusion-xl-base-1.0"
pipe = blocks.init_pipeline(repo_id)
pipe.load_components(torch_dtype=torch.float16, device_map="cuda", trust_remote_code=True)
image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/car.jpg?download=true")
image = image.resize((1024, 1024))
prompt = ["A red car"]
annotation_task = "<REFERRING_EXPRESSION_SEGMENTATION>"
annotation_prompt = ["the car"]
output = pipe(
prompt=prompt,
image=image,
annotation_task=annotation_task,
annotation_prompt=annotation_prompt,
annotation_output_type="mask_image",
num_inference_steps=35,
guidance_scale=7.5,
strength=0.95,
output="images"
)
output[0].save("florence-inpainting.png")
```
## Editing Custom Blocks
By default, custom blocks are saved in your cache directory. Use the `local_dir` argument to download and edit a custom block in a specific folder.
```py
import torch
from diffusers.modular_pipelines import ModularPipelineBlocks, SequentialPipelineBlocks
from diffusers.modular_pipelines.stable_diffusion_xl import INPAINT_BLOCKS
from diffusers.utils import load_image
# Fetch the Florence2 image annotator block that will create our mask
image_annotator_block = ModularPipelineBlocks.from_pretrained("diffusers/florence-2-custom-block", trust_remote_code=True, local_dir="/my-local-folder")
```
Any changes made to the block files in this folder will be reflected when you load the block again.
|