Bria 3 2 pipeline #12010

galbria · 2025-07-29T10:57:41Z

What does this PR do?

Implementing Bria 3.2 pipeline and BriaTransformer2D
issue: here

Before submitting

This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
Did you read the contributor guideline?
Did you read our philosophy doc (important for complex PRs)?
Was this discussed/approved via a GitHub issue or the forum? Please add a link to it if that's the case.
Did you make sure to update the documentation with your changes? Here are the
documentation guidelines, and
here are tips on formatting docstrings.
Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

- Introduced `BriaTransformer2DModel` and `BriaPipeline` for enhanced image generation capabilities. - Updated import structures across various modules to include the new Bria components. - Added utility functions and output classes specific to the Bria pipeline. - Implemented tests for the Bria pipeline to ensure functionality and output integrity.

…bria_3_2_pipeline

SahilCarterr · 2025-07-29T17:21:53Z

Error During Model loading

Code

import torch
from diffusers import BriaPipeline
pipe = BriaPipeline.from_pretrained("briaai/BRIA-3.2", torch_dtype=torch.bfloat16)
pipe.to(device="cuda")
prompt = "A asian girl with red top and blue jeans"
negative_prompt = "Logo,Watermark,Ugly,Morbid,Extra fingers,Poorly drawn hands,Mutation,Blurry,Extra limbs,Gross proportions,Missing arms,Mutated hands,Long neck,Duplicate,Mutilated,Mutilated hands,Poorly drawn face,Deformed,Bad anatomy,Cloned face,Malformed limbs,Missing legs,Too many fingers"

images = pipe(prompt=prompt, negative_prompt=negative_prompt, height=1024, width=1024).images[0]
images

Logs

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
[/tmp/ipython-input-4-3624485676.py](https://localhost:8080/#) in <cell line: 0>()
----> 1 pipe = BriaPipeline.from_pretrained("briaai/BRIA-3.2", torch_dtype=torch.bfloat16)
      2 pipe.to(device="cuda")
      3 prompt = "A asian girl with red top and blue jeans"
      4 negative_prompt = "Logo,Watermark,Ugly,Morbid,Extra fingers,Poorly drawn hands,Mutation,Blurry,Extra limbs,Gross proportions,Missing arms,Mutated hands,Long neck,Duplicate,Mutilated,Mutilated hands,Poorly drawn face,Deformed,Bad anatomy,Cloned face,Malformed limbs,Missing legs,Too many fingers"
      5 

3 frames
[/content/diffusers/src/diffusers/pipelines/pipeline_utils.py](https://localhost:8080/#) in download(cls, pretrained_model_name, **kwargs)
   1536 
   1537             if load_components_from_hub and not trust_remote_code:
-> 1538                 raise ValueError(
   1539                     f"The repository for {pretrained_model_name} contains custom code in {'.py, '.join([os.path.join(k, v) for k, v in custom_components.items()])} which must be executed to correctly "
   1540                     f"load the model. You can inspect the repository content at {', '.join([f'[https://hf.co/{pretrained_model_name}/{k}/{v}.py](https://hf.co/%7Bpretrained_model_name%7D/%7Bk%7D/%7Bv%7D.py)' for k, v in custom_components.items()])}.\n"

ValueError: The repository for briaai/BRIA-3.2 contains custom code in transformer/transformer_bria which must be executed to correctly load the model. You can inspect the repository content at https://hf.co/briaai/BRIA-3.2/transformer/transformer_bria.py.
Please pass the argument `trust_remote_code=True` to allow custom code to be run.

FIX

To fix it transformer_bria.py should be removed and model_index.json should be edited for the same from the model repo on HuggingFace .Always using trust_remote_code=True not efficient for model loading

@galbria

SahilCarterr · 2025-07-29T17:35:17Z

Run make style make quality make fix-copies in order to run checks on github.
@galbria

into bria_3_2_pipeline

galbria · 2025-07-30T11:33:21Z

hey @SahilCarterr i fixed the Huggignface model repo and runt the makes

…dND class for rotary position embedding, and enhance Timestep and TimestepProjEmbeddings classes. Add utility functions for handling negative prompts and generating original sigmas in pipeline_bria.py.

SahilCarterr · 2025-07-30T16:33:27Z

Hey @galbria Checkout the above reviews and edit accordingly

SahilCarterr · 2025-07-30T16:57:00Z

Hey @galbria You need to fix some tests in test_pipeline_bria.py . Below are the error logs

Error Logs

=================================== FAILURES ===================================
________________ BriaPipelineSlowTests.test_bria_inference_bf16 ________________

self = <tests.pipelines.bria.test_pipeline_bria.BriaPipelineSlowTests testMethod=test_bria_inference_bf16>

    def test_bria_inference_bf16(self):
>       pipe = self.pipeline_class.from_pretrained(
            self.repo_id, torch_dtype=torch.bfloat16, text_encoder=None, tokenizer=None
        )

diffusers/tests/pipelines/bria/test_pipeline_bria.py:270: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
/usr/local/lib/python3.11/dist-packages/huggingface_hub/utils/_validators.py:114: in _inner_fn
    return fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^
diffusers/src/diffusers/pipelines/pipeline_utils.py:1093: in from_pretrained
    model = pipeline_class(**init_kwargs)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = BriaPipeline {
  "_class_name": "BriaPipeline",
  "_diffusers_version": "0.35.0.dev0",
  "feature_extractor": [
    nu...ansformer": [
    "diffusers",
    "BriaTransformer2DModel"
  ],
  "vae": [
    "diffusers",
    "AutoencoderKL"
  ]
}

transformer = BriaTransformer2DModel(
  (pos_embed): EmbedND()
  (time_embed): TimestepProjEmbeddings(
    (time_proj): Timesteps()
...((2304,), eps=1e-06, elementwise_affine=False)
  )
  (proj_out): Linear(in_features=2304, out_features=16, bias=True)
)
scheduler = FlowMatchEulerDiscreteScheduler {
  "_class_name": "FlowMatchEulerDiscreteScheduler",
  "_diffusers_version": "0.35.0....beta_sigmas": false,
  "use_dynamic_shifting": true,
  "use_exponential_sigmas": false,
  "use_karras_sigmas": false
}

vae = AutoencoderKL(
  (encoder): Encoder(
    (conv_in): Conv2d(3, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
... Conv2d(8, 8, kernel_size=(1, 1), stride=(1, 1))
  (post_quant_conv): Conv2d(4, 4, kernel_size=(1, 1), stride=(1, 1))
)
text_encoder = None, tokenizer = None, image_encoder = None
feature_extractor = None

    def __init__(
        self,
        transformer: BriaTransformer2DModel,
        scheduler: Union[FlowMatchEulerDiscreteScheduler, KarrasDiffusionSchedulers],
        vae: AutoencoderKL,
        text_encoder: T5EncoderModel,
        tokenizer: T5TokenizerFast,
        image_encoder: CLIPVisionModelWithProjection = None,
        feature_extractor: CLIPImageProcessor = None,
    ):
        self.register_modules(
            vae=vae,
            text_encoder=text_encoder,
            tokenizer=tokenizer,
            transformer=transformer,
            scheduler=scheduler,
            image_encoder=image_encoder,
            feature_extractor=feature_extractor,
        )
    
        # TODO - why different than offical flux (-1)
        self.vae_scale_factor = (
            2 ** (len(self.vae.config.block_out_channels)) if hasattr(self, "vae") and self.vae is not None else 16
        )
        self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
        self.default_sample_size = 64  # due to patchify=> 128,128 => res of 1k,1k
    
        # T5 is senstive to precision so we use the precision used for precompute and cast as needed
>       self.text_encoder = self.text_encoder.to(dtype=T5_PRECISION)
                            ^^^^^^^^^^^^^^^^^^^^
E       AttributeError: 'NoneType' object has no attribute 'to'

diffusers/src/diffusers/pipelines/bria/pipeline_bria.py:189: AttributeError
----------------------------- Captured stderr call -----------------------------
Loading pipeline components...:   0%|          | 0/3 [00:00<?, ?it/s]Loaded vae as AutoencoderKL from `vae` subfolder of briaai/BRIA-3.2.
Loading pipeline components...:  33%|███▎      | 1/3 [00:00<00:00,  6.94it/s]Loaded scheduler as FlowMatchEulerDiscreteScheduler from `scheduler` subfolder of briaai/BRIA-3.2.
Loaded transformer as BriaTransformer2DModel from `transformer` subfolder of briaai/BRIA-3.2.
Loading pipeline components...: 100%|██████████| 3/3 [00:02<00:00,  1.30it/s]
_____________________ BriaPipelineSlowTests.test_to_dtype ______________________

self = <tests.pipelines.bria.test_pipeline_bria.BriaPipelineSlowTests testMethod=test_to_dtype>

    def test_to_dtype(self):
>       components = self.get_dummy_components()
                     ^^^^^^^^^^^^^^^^^^^^^^^^^
E       AttributeError: 'BriaPipelineSlowTests' object has no attribute 'get_dummy_components'

diffusers/tests/pipelines/bria/test_pipeline_bria.py:318: AttributeError
_________________ BriaPipelineNightlyTests.test_bria_inference _________________

self = <tests.pipelines.bria.test_pipeline_bria.BriaPipelineNightlyTests testMethod=test_bria_inference>

    def test_bria_inference(self):
        pipe = BriaPipeline.from_pretrained("briaai/BRIA-3.2", torch_dtype=torch.bfloat16)
        pipe.to(torch_device)
    
        prompt = "a close-up of a smiling cat, high quality, realistic"
        image = pipe(prompt=prompt, num_inference_steps=5, output_type="np").images[0]
    
>       image_slice = image[0, :10, :10, 0].flatten()
                      ^^^^^^^^^^^^^^^^^^^^^
E       IndexError: too many indices for array: array is 3-dimensional, but 4 were indexed

diffusers/tests/pipelines/bria/test_pipeline_bria.py:349: IndexError
----------------------------- Captured stdout call -----------------------------
Using dynamic shift in pipeline with sequence length 4096
----------------------------- Captured stderr call -----------------------------
Loading pipeline components...:   0%|          | 0/5 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/3 [00:00<?, ?it/s]
Loading checkpoint shards:  33%|███▎      | 1/3 [00:02<00:04,  2.01s/it]
Loading checkpoint shards:  67%|██████▋   | 2/3 [00:04<00:02,  2.01s/it]
Loading checkpoint shards: 100%|██████████| 3/3 [00:04<00:00,  1.55s/it]
Loaded text_encoder as T5EncoderModel from `text_encoder` subfolder of briaai/BRIA-3.2.
Loading pipeline components...:  20%|██        | 1/5 [00:06<00:24,  6.15s/it]You set `add_prefix_space`. The tokenizer needs to be converted from the slow tokenizers
Loaded tokenizer as T5TokenizerFast from `tokenizer` subfolder of briaai/BRIA-3.2.
Loading pipeline components...:  40%|████      | 2/5 [00:06<00:08,  2.70s/it]Loaded vae as AutoencoderKL from `vae` subfolder of briaai/BRIA-3.2.
Loading pipeline components...:  60%|██████    | 3/5 [00:06<00:03,  1.52s/it]Loaded scheduler as FlowMatchEulerDiscreteScheduler from `scheduler` subfolder of briaai/BRIA-3.2.
Loaded transformer as BriaTransformer2DModel from `transformer` subfolder of briaai/BRIA-3.2.
Loading pipeline components...: 100%|██████████| 5/5 [00:08<00:00,  1.72s/it]
100%|██████████| 5/5 [00:08<00:00,  1.77s/it]
=========================== short test summary info ============================
FAILED diffusers/tests/pipelines/bria/test_pipeline_bria.py::BriaPipelineSlowTests::test_bria_inference_bf16 - AttributeError: 'NoneType' object has no attribute 'to'
FAILED diffusers/tests/pipelines/bria/test_pipeline_bria.py::BriaPipelineSlowTests::test_to_dtype - AttributeError: 'BriaPipelineSlowTests' object has no attribute 'get_dummy_...
FAILED diffusers/tests/pipelines/bria/test_pipeline_bria.py::BriaPipelineNightlyTests::test_bria_inference - IndexError: too many indices for array: array is 3-dimensional, but 4 were ...

slow test

into bria_3_2_pipeline

galbria · 2025-08-01T14:15:32Z

hey @SahilCarterr fixed the tests.

into bria_3_2_pipeline

linoytsaban · 2025-08-07T07:31:27Z

@bot /style

a-r-r-o-w

Thanks @galbria!

…ng convention

into bria_3_2_pipeline

galbria · 2025-08-15T03:12:09Z

@a-r-r-o-w @SahilCarterr sorry for the additional commit after you have already approved😬
ive just fix the doctree error

and thank you guys for the help!!!

drcgrp · 2025-08-15T17:41:35Z

Please add lora training script from your examples repo here.
Also the import in that repo references a function that doesn't exist in this pipeline atm. Please fix that too

docs/source/en/_toctree.yml

HuggingFaceDocBuilderDev · 2025-08-17T22:32:50Z

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

asomoza · 2025-08-18T03:23:13Z

I tested this model with group offloading so more users can try it without any quality loss, this is the code I used:

import torch

from diffusers import BriaPipeline


onload_device = torch.device("cuda")
offload_device = torch.device("cpu")

pipe = BriaPipeline.from_pretrained("briaai/BRIA-3.2", torch_dtype=torch.float16)
pipe.transformer.enable_group_offload(
    onload_device=onload_device, offload_device=offload_device, offload_type="leaf_level"
)
pipe.to("cuda")

pipe.text_encoder = pipe.text_encoder.to(dtype=torch.float16)
for block in pipe.text_encoder.encoder.block:
    block.layer[-1].DenseReluDense.wo.to(dtype=torch.float32)

if pipe.vae.config.shift_factor == 0:
    pipe.vae.to(dtype=torch.float32)

prompt = "Photorealistic food photography of a stack of fluffy pancakes on a white plate, with maple syrup being poured over them. On top of the pancakes are the words 'BRIA 3.2' in bold, yellow, 3D letters. The background is dark and out of focus."
image = pipe(prompt).images[0]
image.save("bria.png")

So it uses the defaults from the pipeline, however I get this image that seems a lot worse that the one in the repo here, is this expected @galbria?

The one generated with the code seems a lot more fake, like an old 3D render.

DN6

Just one minor comment about the pipeline. Otherwise LGTM 👍🏽

src/diffusers/pipelines/bria/pipeline_bria.py

…uxPipeline, updating imports accordingly.

into bria_3_2_pipeline

galbria · 2025-08-18T16:10:59Z

I tested this model with group offloading so more users can try it without any quality loss, this is the code I used:

import torch

from diffusers import BriaPipeline


onload_device = torch.device("cuda")
offload_device = torch.device("cpu")

pipe = BriaPipeline.from_pretrained("briaai/BRIA-3.2", torch_dtype=torch.float16)
pipe.transformer.enable_group_offload(
    onload_device=onload_device, offload_device=offload_device, offload_type="leaf_level"
)
pipe.to("cuda")

pipe.text_encoder = pipe.text_encoder.to(dtype=torch.float16)
for block in pipe.text_encoder.encoder.block:
    block.layer[-1].DenseReluDense.wo.to(dtype=torch.float32)

if pipe.vae.config.shift_factor == 0:
    pipe.vae.to(dtype=torch.float32)

prompt = "Photorealistic food photography of a stack of fluffy pancakes on a white plate, with maple syrup being poured over them. On top of the pancakes are the words 'BRIA 3.2' in bold, yellow, 3D letters. The background is dark and out of focus."
image = pipe(prompt).images[0]
image.save("bria.png")

So it uses the defaults from the pipeline, however I get this image that seems a lot worse that the one in the repo here, is this expected @galbria?

The one generated with the code seems a lot more fake, like an old 3D render.

I used bfloat16 instead of float16 and got a better result. Does it help?

src/diffusers/pipelines/bria/pipeline_bria.py

src/diffusers/models/transformers/transformer_bria.py

asomoza · 2025-08-19T11:07:49Z

I used bfloat16 instead of float16 and got a better result. Does it help?

@galbria I was just commenting the difference with the example image, if you're good with the result and it's expected, it's your decision. My comment is definitely not a merge blocker.

into bria_3_2_pipeline

galbria · 2025-08-19T12:52:45Z

@SahilCarterr @asomoza @a-r-r-o-w @yiyixuxu What else should I do so we can merge it?

a-r-r-o-w

Just one last question

src/diffusers/pipelines/bria/pipeline_bria.py

… for better result

into bria_3_2_pipeline

galbria added 6 commits July 28, 2025 08:37

with working tests

7808ee0

style and quality pass

c58267e

Merge branch 'main' of https://github.com/huggingface/diffusers into …

4208a09

…bria_3_2_pipeline

adding docs

7c1cf7e

add to overview

92671ab

galbria mentioned this pull request Jul 29, 2025

Add Bria 3.2 pipeline - next-generation commercial-ready text-to-image model #11978

Open

2 tasks

galbria and others added 3 commits July 30, 2025 14:29

Merge branch 'main' into bria_3_2_pipeline

649767f

fixes from "make fix-copies"

be29631

Merge branch 'bria_3_2_pipeline' of https://github.com/galbria/diffusers

eb1a3f4

into bria_3_2_pipeline

Refactor transformer_bria.py and pipeline_bria.py: Introduce new Embe…

88c8e42

…dND class for rotary position embedding, and enhance Timestep and TimestepProjEmbeddings classes. Add utility functions for handling negative prompts and generating original sigmas in pipeline_bria.py.

galbria and others added 5 commits July 31, 2025 08:40

remove redundent and duplicates tests and fix bf16

6cefe44

slow test

Merge branch 'main' into bria_3_2_pipeline

770e162

style fixes

f27d122

Merge branch 'bria_3_2_pipeline' of https://github.com/galbria/diffusers

f48d072

into bria_3_2_pipeline

Merge branch 'main' into bria_3_2_pipeline

e7b6db5

galbria and others added 5 commits August 3, 2025 09:40

Merge branch 'main' into bria_3_2_pipeline

7091624

Merge branch 'main' into bria_3_2_pipeline

7b4b8d1

small doc update

a423221

Merge branch 'bria_3_2_pipeline' of https://github.com/galbria/diffusers

63c9078

into bria_3_2_pipeline

Merge branch 'main' into bria_3_2_pipeline

6f656e7

linoytsaban requested review from asomoza and yiyixuxu August 7, 2025 07:28

a-r-r-o-w approved these changes Aug 15, 2025

View reviewed changes

galbria added 2 commits August 15, 2025 03:07

Update Bria model reference in documentation to reflect new file nami…

d7dd8b3

…ng convention

Merge branch 'bria_3_2_pipeline' of https://github.com/galbria/diffusers

4dd04ba

into bria_3_2_pipeline

galbria requested a review from a-r-r-o-w August 15, 2025 03:12

a-r-r-o-w reviewed Aug 17, 2025

View reviewed changes

docs/source/en/_toctree.yml Outdated Show resolved Hide resolved

Update docs/source/en/_toctree.yml

36b0133

Merge branch 'main' into bria_3_2_pipeline

51a3bdc

DN6 reviewed Aug 18, 2025

View reviewed changes

src/diffusers/pipelines/bria/pipeline_bria.py Outdated Show resolved Hide resolved

galbria and others added 3 commits August 18, 2025 15:14

Refactor BriaPipeline to inherit from DiffusionPipeline instead of Fl…

9c6d9dd

…uxPipeline, updating imports accordingly.

Merge branch 'bria_3_2_pipeline' of https://github.com/galbria/diffusers

dd9dfd8

into bria_3_2_pipeline

Merge branch 'main' into bria_3_2_pipeline

3dab480

SahilCarterr reviewed Aug 19, 2025

View reviewed changes

SahilCarterr approved these changes Aug 19, 2025

View reviewed changes

galbria and others added 3 commits August 19, 2025 12:29

move the __call__ func to the end of file

f702292

Merge branch 'bria_3_2_pipeline' of https://github.com/galbria/diffusers

4331603

into bria_3_2_pipeline

Merge branch 'main' into bria_3_2_pipeline

1456f5d

a-r-r-o-w approved these changes Aug 20, 2025

View reviewed changes

src/diffusers/pipelines/bria/pipeline_bria.py Outdated Show resolved Hide resolved

galbria and others added 5 commits August 20, 2025 08:37

Update BriaPipeline example to use bfloat16 for precision sensitivity…

4430fe9

… for better result

Merge branch 'bria_3_2_pipeline' of https://github.com/galbria/diffusers

d20b9d8

into bria_3_2_pipeline

Merge branch 'main' into bria_3_2_pipeline

f9d5e46

make style && make quality && make fix-copiessource

3b847ab

Merge branch 'bria_3_2_pipeline' of https://github.com/galbria/diffusers

135a3aa

into bria_3_2_pipeline

a-r-r-o-w merged commit 7993be9 into huggingface:main Aug 20, 2025
11 checks passed

Bria 3 2 pipeline #12010

Bria 3 2 pipeline #12010

Uh oh!

Conversation

galbria commented Jul 29, 2025

What does this PR do?

Before submitting

Who can review?

Uh oh!

SahilCarterr commented Jul 29, 2025

Error During Model loading

FIX

Uh oh!

SahilCarterr commented Jul 29, 2025

Uh oh!

galbria commented Jul 30, 2025

Uh oh!

SahilCarterr commented Jul 30, 2025

Uh oh!

SahilCarterr commented Jul 30, 2025

Uh oh!

galbria commented Aug 1, 2025

Uh oh!

linoytsaban commented Aug 7, 2025

Uh oh!

a-r-r-o-w left a comment

Choose a reason for hiding this comment

Uh oh!

galbria commented Aug 15, 2025

Uh oh!

drcgrp commented Aug 15, 2025

Uh oh!

Uh oh!

HuggingFaceDocBuilderDev commented Aug 17, 2025

Uh oh!

asomoza commented Aug 18, 2025

Uh oh!

DN6 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

galbria commented Aug 18, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

asomoza commented Aug 19, 2025

Uh oh!

galbria commented Aug 19, 2025

Uh oh!

a-r-r-o-w left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!