├── .gitignore ├── LICENSE ├── README.md ├── controlnet ├── .bentoignore ├── README.md ├── example-image.png ├── hf-logo.png ├── requirements.txt └── service.py ├── flux-timestep-distilled ├── README.md ├── requirements.txt └── service.py ├── sd3-medium ├── .bentoignore ├── README.md ├── bentofile.yaml ├── requirements.txt └── service.py ├── sd3.5-large-turbo ├── .bentoignore ├── README.md ├── bentofile.yaml ├── requirements.txt └── service.py ├── sd3.5-large ├── .bentoignore ├── README.md ├── bentofile.yaml ├── requirements.txt └── service.py ├── sdxl-lightning ├── .bentoignore ├── README.md ├── bentofile.yaml ├── requirements.txt └── service.py └── sdxl-turbo ├── .bentoignore ├── README.md ├── requirements.txt └── service.py /.gitignore: -------------------------------------------------------------------------------- 1 | # Byte-compiled / optimized / DLL files 2 | __pycache__/ 3 | *.py[cod] 4 | *$py.class 5 | 6 | # C extensions 7 | *.so 8 | 9 | # Distribution / packaging 10 | .Python 11 | build/ 12 | develop-eggs/ 13 | dist/ 14 | downloads/ 15 | eggs/ 16 | .eggs/ 17 | lib/ 18 | lib64/ 19 | parts/ 20 | sdist/ 21 | var/ 22 | wheels/ 23 | pip-wheel-metadata/ 24 | share/python-wheels/ 25 | *.egg-info/ 26 | .installed.cfg 27 | *.egg 28 | MANIFEST 29 | 30 | # PyInstaller 31 | # Usually these files are written by a python script from a template 32 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 33 | *.manifest 34 | *.spec 35 | 36 | # Installer logs 37 | pip-log.txt 38 | pip-delete-this-directory.txt 39 | 40 | # Unit test / coverage reports 41 | htmlcov/ 42 | .tox/ 43 | .nox/ 44 | .coverage 45 | .coverage.* 46 | .cache 47 | nosetests.xml 48 | coverage.xml 49 | *.cover 50 | *.py,cover 51 | .hypothesis/ 52 | .pytest_cache/ 53 | 54 | # Translations 55 | *.mo 56 | *.pot 57 | 58 | # Django stuff: 59 | *.log 60 | local_settings.py 61 | db.sqlite3 62 | db.sqlite3-journal 63 | 64 | # Flask stuff: 65 | instance/ 66 | .webassets-cache 67 | 68 | # Scrapy stuff: 69 | .scrapy 70 | 71 | # Sphinx documentation 72 | docs/_build/ 73 | 74 | # PyBuilder 75 | target/ 76 | 77 | # Jupyter Notebook 78 | .ipynb_checkpoints 79 | 80 | # IPython 81 | profile_default/ 82 | ipython_config.py 83 | 84 | # pyenv 85 | .python-version 86 | 87 | # pipenv 88 | # According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control. 89 | # However, in case of collaboration, if having platform-specific dependencies or dependencies 90 | # having no cross-platform support, pipenv may install dependencies that don't work, or not 91 | # install all needed dependencies. 92 | #Pipfile.lock 93 | 94 | # PEP 582; used by e.g. github.com/David-OConnor/pyflow 95 | __pypackages__/ 96 | 97 | # Celery stuff 98 | celerybeat-schedule 99 | celerybeat.pid 100 | 101 | # SageMath parsed files 102 | *.sage.py 103 | 104 | # Environments 105 | .env 106 | .venv 107 | env/ 108 | venv/ 109 | ENV/ 110 | env.bak/ 111 | venv.bak/ 112 | 113 | # Spyder project settings 114 | .spyderproject 115 | .spyproject 116 | 117 | # Rope project settings 118 | .ropeproject 119 | 120 | # mkdocs documentation 121 | /site 122 | 123 | # mypy 124 | .mypy_cache/ 125 | .dmypy.json 126 | dmypy.json 127 | 128 | # Pyre type checker 129 | .pyre/ 130 | bazel-* 131 | 132 | package-lock.json 133 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | Apache License 2 | Version 2.0, January 2004 3 | http://www.apache.org/licenses/ 4 | 5 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 6 | 7 | 1. Definitions. 8 | 9 | "License" shall mean the terms and conditions for use, reproduction, 10 | and distribution as defined by Sections 1 through 9 of this document. 11 | 12 | "Licensor" shall mean the copyright owner or entity authorized by 13 | the copyright owner that is granting the License. 14 | 15 | "Legal Entity" shall mean the union of the acting entity and all 16 | other entities that control, are controlled by, or are under common 17 | control with that entity. For the purposes of this definition, 18 | "control" means (i) the power, direct or indirect, to cause the 19 | direction or management of such entity, whether by contract or 20 | otherwise, or (ii) ownership of fifty percent (50%) or more of the 21 | outstanding shares, or (iii) beneficial ownership of such entity. 22 | 23 | "You" (or "Your") shall mean an individual or Legal Entity 24 | exercising permissions granted by this License. 25 | 26 | "Source" form shall mean the preferred form for making modifications, 27 | including but not limited to software source code, documentation 28 | source, and configuration files. 29 | 30 | "Object" form shall mean any form resulting from mechanical 31 | transformation or translation of a Source form, including but 32 | not limited to compiled object code, generated documentation, 33 | and conversions to other media types. 34 | 35 | "Work" shall mean the work of authorship, whether in Source or 36 | Object form, made available under the License, as indicated by a 37 | copyright notice that is included in or attached to the work 38 | (an example is provided in the Appendix below). 39 | 40 | "Derivative Works" shall mean any work, whether in Source or Object 41 | form, that is based on (or derived from) the Work and for which the 42 | editorial revisions, annotations, elaborations, or other modifications 43 | represent, as a whole, an original work of authorship. For the purposes 44 | of this License, Derivative Works shall not include works that remain 45 | separable from, or merely link (or bind by name) to the interfaces of, 46 | the Work and Derivative Works thereof. 47 | 48 | "Contribution" shall mean any work of authorship, including 49 | the original version of the Work and any modifications or additions 50 | to that Work or Derivative Works thereof, that is intentionally 51 | submitted to Licensor for inclusion in the Work by the copyright owner 52 | or by an individual or Legal Entity authorized to submit on behalf of 53 | the copyright owner. For the purposes of this definition, "submitted" 54 | means any form of electronic, verbal, or written communication sent 55 | to the Licensor or its representatives, including but not limited to 56 | communication on electronic mailing lists, source code control systems, 57 | and issue tracking systems that are managed by, or on behalf of, the 58 | Licensor for the purpose of discussing and improving the Work, but 59 | excluding communication that is conspicuously marked or otherwise 60 | designated in writing by the copyright owner as "Not a Contribution." 61 | 62 | "Contributor" shall mean Licensor and any individual or Legal Entity 63 | on behalf of whom a Contribution has been received by Licensor and 64 | subsequently incorporated within the Work. 65 | 66 | 2. Grant of Copyright License. Subject to the terms and conditions of 67 | this License, each Contributor hereby grants to You a perpetual, 68 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 69 | copyright license to reproduce, prepare Derivative Works of, 70 | publicly display, publicly perform, sublicense, and distribute the 71 | Work and such Derivative Works in Source or Object form. 72 | 73 | 3. Grant of Patent License. Subject to the terms and conditions of 74 | this License, each Contributor hereby grants to You a perpetual, 75 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 76 | (except as stated in this section) patent license to make, have made, 77 | use, offer to sell, sell, import, and otherwise transfer the Work, 78 | where such license applies only to those patent claims licensable 79 | by such Contributor that are necessarily infringed by their 80 | Contribution(s) alone or by combination of their Contribution(s) 81 | with the Work to which such Contribution(s) was submitted. If You 82 | institute patent litigation against any entity (including a 83 | cross-claim or counterclaim in a lawsuit) alleging that the Work 84 | or a Contribution incorporated within the Work constitutes direct 85 | or contributory patent infringement, then any patent licenses 86 | granted to You under this License for that Work shall terminate 87 | as of the date such litigation is filed. 88 | 89 | 4. Redistribution. You may reproduce and distribute copies of the 90 | Work or Derivative Works thereof in any medium, with or without 91 | modifications, and in Source or Object form, provided that You 92 | meet the following conditions: 93 | 94 | (a) You must give any other recipients of the Work or 95 | Derivative Works a copy of this License; and 96 | 97 | (b) You must cause any modified files to carry prominent notices 98 | stating that You changed the files; and 99 | 100 | (c) You must retain, in the Source form of any Derivative Works 101 | that You distribute, all copyright, patent, trademark, and 102 | attribution notices from the Source form of the Work, 103 | excluding those notices that do not pertain to any part of 104 | the Derivative Works; and 105 | 106 | (d) If the Work includes a "NOTICE" text file as part of its 107 | distribution, then any Derivative Works that You distribute must 108 | include a readable copy of the attribution notices contained 109 | within such NOTICE file, excluding those notices that do not 110 | pertain to any part of the Derivative Works, in at least one 111 | of the following places: within a NOTICE text file distributed 112 | as part of the Derivative Works; within the Source form or 113 | documentation, if provided along with the Derivative Works; or, 114 | within a display generated by the Derivative Works, if and 115 | wherever such third-party notices normally appear. The contents 116 | of the NOTICE file are for informational purposes only and 117 | do not modify the License. You may add Your own attribution 118 | notices within Derivative Works that You distribute, alongside 119 | or as an addendum to the NOTICE text from the Work, provided 120 | that such additional attribution notices cannot be construed 121 | as modifying the License. 122 | 123 | You may add Your own copyright statement to Your modifications and 124 | may provide additional or different license terms and conditions 125 | for use, reproduction, or distribution of Your modifications, or 126 | for any such Derivative Works as a whole, provided Your use, 127 | reproduction, and distribution of the Work otherwise complies with 128 | the conditions stated in this License. 129 | 130 | 5. Submission of Contributions. Unless You explicitly state otherwise, 131 | any Contribution intentionally submitted for inclusion in the Work 132 | by You to the Licensor shall be under the terms and conditions of 133 | this License, without any additional terms or conditions. 134 | Notwithstanding the above, nothing herein shall supersede or modify 135 | the terms of any separate license agreement you may have executed 136 | with Licensor regarding such Contributions. 137 | 138 | 6. Trademarks. This License does not grant permission to use the trade 139 | names, trademarks, service marks, or product names of the Licensor, 140 | except as required for reasonable and customary use in describing the 141 | origin of the Work and reproducing the content of the NOTICE file. 142 | 143 | 7. Disclaimer of Warranty. Unless required by applicable law or 144 | agreed to in writing, Licensor provides the Work (and each 145 | Contributor provides its Contributions) on an "AS IS" BASIS, 146 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or 147 | implied, including, without limitation, any warranties or conditions 148 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A 149 | PARTICULAR PURPOSE. You are solely responsible for determining the 150 | appropriateness of using or redistributing the Work and assume any 151 | risks associated with Your exercise of permissions under this License. 152 | 153 | 8. Limitation of Liability. In no event and under no legal theory, 154 | whether in tort (including negligence), contract, or otherwise, 155 | unless required by applicable law (such as deliberate and grossly 156 | negligent acts) or agreed to in writing, shall any Contributor be 157 | liable to You for damages, including any direct, indirect, special, 158 | incidental, or consequential damages of any character arising as a 159 | result of this License or out of the use or inability to use the 160 | Work (including but not limited to damages for loss of goodwill, 161 | work stoppage, computer failure or malfunction, or any and all 162 | other commercial damages or losses), even if such Contributor 163 | has been advised of the possibility of such damages. 164 | 165 | 9. Accepting Warranty or Additional Liability. While redistributing 166 | the Work or Derivative Works thereof, You may choose to offer, 167 | and charge a fee for, acceptance of support, warranty, indemnity, 168 | or other liability obligations and/or rights consistent with this 169 | License. However, in accepting such obligations, You may act only 170 | on Your own behalf and on Your sole responsibility, not on behalf 171 | of any other Contributor, and only if You agree to indemnify, 172 | defend, and hold each Contributor harmless for any liability 173 | incurred by, or claims asserted against, such Contributor by reason 174 | of your accepting any such warranty or additional liability. 175 | 176 | END OF TERMS AND CONDITIONS 177 | 178 | APPENDIX: How to apply the Apache License to your work. 179 | 180 | To apply the Apache License to your work, attach the following 181 | boilerplate notice, with the fields enclosed by brackets "[]" 182 | replaced with your own identifying information. (Don't include 183 | the brackets!) The text should be enclosed in the appropriate 184 | comment syntax for the file format. We also recommend that a 185 | file or class name and description of purpose be included on the 186 | same "printed page" as the copyright notice for easier 187 | identification within third-party archives. 188 | 189 | Copyright [yyyy] [name of copyright owner] 190 | 191 | Licensed under the Apache License, Version 2.0 (the "License"); 192 | you may not use this file except in compliance with the License. 193 | You may obtain a copy of the License at 194 | 195 | http://www.apache.org/licenses/LICENSE-2.0 196 | 197 | Unless required by applicable law or agreed to in writing, software 198 | distributed under the License is distributed on an "AS IS" BASIS, 199 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 200 | See the License for the specific language governing permissions and 201 | limitations under the License. 202 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 |
2 |

Self-host Diffusion Models with BentoML

3 |
4 | 5 | This repository contains a series of BentoML example projects, demonstrating how to deploy different models in [the Stable Diffusion (SD) family](https://huggingface.co/models?other=stable-diffusion), which is specialized in generating and manipulating images or video clips based on text prompts. 6 | 7 | See [here](https://docs.bentoml.com/en/latest/examples/overview.html) for a full list of BentoML example projects. 8 | 9 | The following guide uses SDXL Turbo as an example. 10 | 11 | ## Prerequisites 12 | 13 | If you want to test the Service locally, we recommend you use an Nvidia GPU with at least 12GB VRAM. 14 | 15 | ## Install dependencies 16 | 17 | ```bash 18 | git clone https://github.com/bentoml/BentoDiffusion.git 19 | cd BentoDiffusion/sdxl-turbo 20 | 21 | # Recommend Python 3.11 22 | pip install -r requirements.txt 23 | ``` 24 | 25 | ## Run the BentoML Service 26 | 27 | We have defined a BentoML Service in `service.py`. Run `bentoml serve` in your project directory to start the Service. 28 | 29 | ```bash 30 | $ bentoml serve 31 | 32 | 2024-01-18T18:31:49+0800 [INFO] [cli] Starting production HTTP BentoServer from "service:SDXLTurboService" listening on http://localhost:3000 (Press CTRL+C to quit) 33 | Loading pipeline components...: 100% 34 | ``` 35 | 36 | The server is now active at [http://localhost:3000](http://localhost:3000/). You can interact with it using the Swagger UI or in other different ways. 37 | 38 | CURL 39 | 40 | ```bash 41 | curl -X 'POST' \ 42 | 'http://localhost:3000/txt2img' \ 43 | -H 'accept: image/*' \ 44 | -H 'Content-Type: application/json' \ 45 | -d '{ 46 | "prompt": "A cinematic shot of a baby racoon wearing an intricate italian priest robe.", 47 | "num_inference_steps": 1, 48 | "guidance_scale": 0 49 | }' 50 | ``` 51 | 52 | Python client 53 | 54 | ```python 55 | import bentoml 56 | 57 | with bentoml.SyncHTTPClient("http://localhost:3000") as client: 58 | result = client.txt2img( 59 | prompt="A cinematic shot of a baby racoon wearing an intricate italian priest robe.", 60 | num_inference_steps=1, 61 | guidance_scale=0.0 62 | ) 63 | ``` 64 | 65 | For detailed explanations of the Service code, see [Stable Diffusion XL Turbo](https://docs.bentoml.com/en/latest/use-cases/diffusion-models/sdxl-turbo.html). 66 | 67 | ## Deploy to BentoCloud 68 | 69 | After the Service is ready, you can deploy the application to BentoCloud for better management and scalability. [Sign up](https://www.bentoml.com/) if you haven't got a BentoCloud account. 70 | 71 | Make sure you have [logged in to BentoCloud](https://docs.bentoml.com/en/latest/scale-with-bentocloud/manage-api-tokens.html). 72 | 73 | ```bash 74 | bentoml cloud login 75 | ``` 76 | 77 | Deploy it to BentoCloud. 78 | 79 | ```bash 80 | bentoml deploy 81 | ``` 82 | 83 | Once the application is up and running on BentoCloud, you can access it via the exposed URL. 84 | 85 | **Note**: For custom deployment in your own infrastructure, use [BentoML to generate an OCI-compliant image](https://docs.bentoml.com/en/latest/get-started/packaging-for-deployment.html). 86 | 87 | 88 | ## Choose another diffusion model 89 | 90 | To deploy a different diffusion model, go to the corresponding subdirectories of this repository. 91 | 92 | - [FLUX.1](flux-timestep-distilled/) 93 | - [Stable Diffusion 3 Medium](sd3-medium/) 94 | - [Stable Diffusion 3.5 Large Turbo](sd3.5-large-turbo/) 95 | - [Stable Diffusion 3.5 Large](sd3.5-large/) 96 | - [Stable Diffusion XL Lightning](sdxl-lightning/) 97 | - [Stable Diffusion XL Turbo](sdxl-turbo/) 98 | - [ControlNet](controlnet/) -------------------------------------------------------------------------------- /controlnet/.bentoignore: -------------------------------------------------------------------------------- 1 | __pycache__/ 2 | venv/ 3 | -------------------------------------------------------------------------------- /controlnet/README.md: -------------------------------------------------------------------------------- 1 |
2 |

Serving SDXL and ControlNet with BentoML

3 |
4 | 5 | ControlNet is a model designed to control image diffusion processes by conditioning them with additional input images, such as canny edges, user sketches, human poses, depth maps, and more. This allows for greater control over image generation by guiding the model with specific inputs, making it easier to generate targeted images. 6 | 7 | This is a BentoML example project, demonstrating how to build an image generation inference API server, using the [SDXL model](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0) and [the ControlNet model](https://huggingface.co/diffusers/controlnet-canny-sdxl-1.0). 8 | 9 | See [here](https://docs.bentoml.com/en/latest/examples/overview.html) for a full list of BentoML example projects. 10 | 11 | ## Prerequisites 12 | 13 | If you want to test the Service locally, we recommend you use an Nvidia GPU with at least 12GB VRAM. 14 | 15 | ## Install dependencies 16 | 17 | ```bash 18 | git clone https://github.com/bentoml/BentoDiffusion.git 19 | cd BentoDiffusion/controlnet 20 | 21 | # Recommend Python 3.11 22 | pip install -r requirements.txt 23 | ``` 24 | 25 | ## Run the BentoML Service 26 | 27 | We have defined a BentoML Service in `service.py`. Run `bentoml serve` in your project directory to start the Service. 28 | 29 | ```bash 30 | $ bentoml serve 31 | 32 | 2024-01-18T09:43:40+0800 [INFO] [cli] Prometheus metrics for HTTP BentoServer from "service:APIService" can be accessed at http://localhost:3000/metrics. 33 | 2024-01-18T09:43:41+0800 [INFO] [cli] Starting production HTTP BentoServer from "service:APIService" listening on http://localhost:3000 (Press CTRL+C to quit) 34 | ``` 35 | 36 | The server is now active at [http://localhost:3000](http://localhost:3000/). You can interact with it using the Swagger UI or in other different ways. 37 | 38 | CURL 39 | 40 | ```bash 41 | curl -X 'POST' \ 42 | 'http://localhost:3000/generate' \ 43 | -H 'accept: image/*' \ 44 | -H 'Content-Type: multipart/form-data' \ 45 | -F 'image=@example-image.png;type=image/png' \ 46 | -F 'prompt=A young man walking in a park, wearing jeans.' \ 47 | -F 'negative_prompt=ugly, disfigured, ill-structured, low resolution' \ 48 | -F 'controlnet_conditioning_scale=0.5' \ 49 | -F 'num_inference_steps=25' 50 | ``` 51 | 52 | Python client 53 | 54 | ```python 55 | import bentoml 56 | from pathlib import Path 57 | 58 | with bentoml.SyncHTTPClient("http://localhost:3000") as client: 59 | result = client.generate( 60 | image=Path("example-image.png"), 61 | prompt="A young man walking in a park, wearing jeans.", 62 | negative_prompt="ugly, disfigured, ill-structure, low resolution", 63 | controlnet_conditioning_scale=0.5, 64 | num_inference_steps=25, 65 | ) 66 | ``` 67 | 68 | For detailed explanations of the Service code, see [ControlNet](https://docs.bentoml.com/en/latest/examples/controlnet.html). 69 | 70 | ## Deploy to BentoCloud 71 | 72 | After the Service is ready, you can deploy the application to BentoCloud for better management and scalability. [Sign up](https://www.bentoml.com/) if you haven't got a BentoCloud account. 73 | 74 | Make sure you have [logged in to BentoCloud](https://docs.bentoml.com/en/latest/scale-with-bentocloud/manage-api-tokens.html). 75 | 76 | ```bash 77 | bentoml cloud login 78 | ``` 79 | 80 | Deploy it to BentoCloud. 81 | 82 | ```bash 83 | bentoml deploy 84 | ``` 85 | 86 | Once the application is up and running on BentoCloud, you can access it via the exposed URL. 87 | 88 | **Note**: For custom deployment in your own infrastructure, use [BentoML to generate an OCI-compliant image](https://docs.bentoml.com/en/latest/get-started/packaging-for-deployment.html). 89 | -------------------------------------------------------------------------------- /controlnet/example-image.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/bentoml/BentoDiffusion/41a42d606d73697ed9a7bb66c0d6fc879875db18/controlnet/example-image.png -------------------------------------------------------------------------------- /controlnet/hf-logo.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/bentoml/BentoDiffusion/41a42d606d73697ed9a7bb66c0d6fc879875db18/controlnet/hf-logo.png -------------------------------------------------------------------------------- /controlnet/requirements.txt: -------------------------------------------------------------------------------- 1 | accelerate==0.31.0 2 | bentoml==1.4.11 3 | diffusers==0.29.0 4 | opencv-python==4.10.0.84 5 | pillow==10.3.0 6 | torch==2.6.0 7 | transformers==4.48.0 8 | -------------------------------------------------------------------------------- /controlnet/service.py: -------------------------------------------------------------------------------- 1 | from __future__ import annotations 2 | 3 | import typing as t 4 | 5 | import numpy as np 6 | import PIL 7 | from PIL.Image import Image as PIL_Image 8 | 9 | import bentoml 10 | 11 | CONTROLNET_MODEL_ID = "diffusers/controlnet-canny-sdxl-1.0" 12 | VAE_MODEL_ID = "madebyollin/sdxl-vae-fp16-fix" 13 | BASE_MODEL_ID = "stabilityai/stable-diffusion-xl-base-1.0" 14 | 15 | 16 | my_image = bentoml.images.PythonImage(python_version='3.11', distro='debian') \ 17 | .system_packages("ffmpeg") \ 18 | .requirements_file("requirements.txt") 19 | 20 | 21 | @bentoml.service( 22 | image=my_image, 23 | traffic={"timeout": 600}, 24 | workers=1, 25 | labels={'owner': 'bentoml-team', 'project': 'gallery'}, 26 | resources={ 27 | "gpu": 1, 28 | "gpu_type": "nvidia-l4", 29 | } 30 | ) 31 | class ControlNet: 32 | controlnet_path = bentoml.models.HuggingFaceModel(CONTROLNET_MODEL_ID) 33 | vae_path = bentoml.models.HuggingFaceModel(VAE_MODEL_ID) 34 | base_path = bentoml.models.HuggingFaceModel(BASE_MODEL_ID) 35 | 36 | def __init__(self) -> None: 37 | 38 | import torch 39 | from diffusers import StableDiffusionXLControlNetPipeline, ControlNetModel, AutoencoderKL 40 | 41 | if torch.cuda.is_available(): 42 | self.device = "cuda" 43 | self.dtype = torch.float16 44 | else: 45 | self.device = "cpu" 46 | self.dtype = torch.float32 47 | 48 | self.controlnet = ControlNetModel.from_pretrained( 49 | self.controlnet_path, 50 | torch_dtype=self.dtype, 51 | ) 52 | 53 | self.vae = AutoencoderKL.from_pretrained( 54 | self.vae_path, 55 | torch_dtype=self.dtype, 56 | ) 57 | 58 | self.pipe = StableDiffusionXLControlNetPipeline.from_pretrained( 59 | self.base_path, 60 | controlnet=self.controlnet, 61 | vae=self.vae, 62 | torch_dtype=self.dtype 63 | ).to(self.device) 64 | 65 | @bentoml.api 66 | def generate( 67 | self, 68 | image: PIL_Image, 69 | prompt: str, 70 | negative_prompt: t.Optional[str] = None, 71 | controlnet_conditioning_scale: t.Optional[float] = 1.0, 72 | num_inference_steps: t.Optional[int] = 50, 73 | guidance_scale: t.Optional[float] = 5.0, 74 | ) -> PIL_Image: 75 | import cv2 76 | 77 | if controlnet_conditioning_scale is None: 78 | controlnet_conditioning_scale = 1.0 79 | 80 | if num_inference_steps is None: 81 | num_inference_steps = 50 82 | 83 | if guidance_scale is None: 84 | guidance_scale = 5.0 85 | 86 | arr = np.array(image) 87 | arr = cv2.Canny(arr, 100, 200) 88 | arr = arr[:, :, None] 89 | arr = np.concatenate([arr, arr, arr], axis=2) 90 | image = PIL.Image.fromarray(arr) 91 | return self.pipe( 92 | prompt, 93 | image=image, 94 | negative_prompt=negative_prompt, 95 | controlnet_conditioning_scale=controlnet_conditioning_scale, 96 | num_inference_steps=num_inference_steps, 97 | guidance_scale=guidance_scale, 98 | ).to_tuple()[0][0] 99 | -------------------------------------------------------------------------------- /flux-timestep-distilled/README.md: -------------------------------------------------------------------------------- 1 |
2 |

Serving FLUX.1 models with BentoML

3 |
4 | 5 | This is a BentoML example project, demonstrating how to build an image generation inference API server using the [FLUX.1-schnell](https://huggingface.co/black-forest-labs/FLUX.1-schnell), a 12 billion parameter rectified flow transformer capable of generating images from text descriptions. 6 | 7 | See [here](https://docs.bentoml.com/en/latest/examples/overview.html) for a full list of BentoML example projects. 8 | 9 | ## Prerequisites 10 | 11 | To run the Service locally, we recommend you use an Nvidia GPU with at least 32G VRAM. 12 | 13 | ## Install dependencies 14 | 15 | ```bash 16 | git clone https://github.com/bentoml/BentoDiffusion.git 17 | cd BentoDiffusion/flux-timestep-distilled 18 | 19 | # Recommend Python 3.11 20 | pip install -r requirements.txt 21 | ``` 22 | 23 | ## Run the BentoML Service 24 | 25 | We have defined a BentoML Service in `service.py`. Run `bentoml serve` in your project directory to start the Service. 26 | 27 | ```bash 28 | $ bentoml serve 29 | 30 | 2024-01-18T18:31:49+0800 [INFO] [cli] Starting production HTTP BentoServer from "service:FluxTimestepDistilled" listening on http://localhost:3000 (Press CTRL+C to quit) 31 | Loading pipeline components...: 100% 32 | ``` 33 | 34 | The server is now active at [http://localhost:3000](http://localhost:3000/). You can interact with it using the Swagger UI or in other different ways. 35 | 36 | CURL 37 | 38 | ```bash 39 | curl -X 'POST' \ 40 | 'http://localhost:3000/txt2img' \ 41 | -H 'accept: image/*' \ 42 | -H 'Content-Type: application/json' \ 43 | -d '{ 44 | "prompt": "A cinematic shot of a baby racoon wearing an intricate italian priest robe.", 45 | }' 46 | ``` 47 | 48 | BentoML client 49 | 50 | ```python 51 | import bentoml 52 | 53 | with bentoml.SyncHTTPClient("http://localhost:3000") as client: 54 | result = client.txt2img( 55 | prompt="A cinematic shot of a baby racoon wearing an intricate italian priest robe.", 56 | ) 57 | ``` 58 | 59 | ## Deploy to BentoCloud 60 | 61 | After the Service is ready, you can deploy the application to BentoCloud for better management and scalability. [Sign up](https://www.bentoml.com/) if you haven't got a BentoCloud account. 62 | 63 | Make sure you have [logged in to BentoCloud](https://docs.bentoml.com/en/latest/scale-with-bentocloud/manage-api-tokens.html). 64 | 65 | ```bash 66 | bentoml cloud login 67 | ``` 68 | 69 | Deploy it to BentoCloud. 70 | 71 | ```bash 72 | bentoml deploy 73 | ``` 74 | 75 | Once the application is up and running on BentoCloud, you can access it via the exposed URL. 76 | 77 | **Note**: For custom deployment in your own infrastructure, use [BentoML to generate an OCI-compliant image](https://docs.bentoml.com/en/latest/get-started/packaging-for-deployment.html). 78 | -------------------------------------------------------------------------------- /flux-timestep-distilled/requirements.txt: -------------------------------------------------------------------------------- 1 | accelerate==1.6.0 2 | bentoml==1.4.11 3 | diffusers==0.33.1 4 | pillow==11.2.1 5 | protobuf==6.30.1 6 | sentencepiece==0.2.0 7 | torch==2.6.0 8 | transformers==4.51.0 9 | -------------------------------------------------------------------------------- /flux-timestep-distilled/service.py: -------------------------------------------------------------------------------- 1 | from __future__ import annotations 2 | 3 | import bentoml 4 | from PIL.Image import Image 5 | 6 | MODEL_ID = "black-forest-labs/FLUX.1-schnell" 7 | 8 | sample_prompt = "A girl smiling" 9 | 10 | 11 | @bentoml.service( 12 | name="bento-flux-timestep-distilled-service", 13 | image=bentoml.images.Image(python_version="3.11").requirements_file("requirements.txt"), 14 | traffic={"timeout": 300}, 15 | envs=[{"name": "HF_TOKEN"}], 16 | resources={"gpu": 1, "gpu_type": "nvidia-a100-80gb"}, 17 | ) 18 | class FluxTimestepDistilled: 19 | @bentoml.on_startup 20 | def setup_pipeline(self) -> None: 21 | import torch 22 | from diffusers import FluxPipeline 23 | 24 | self.pipe = FluxPipeline.from_pretrained( 25 | MODEL_ID, 26 | torch_dtype=torch.bfloat16, 27 | use_safetensors=True, 28 | ) 29 | self.pipe.to("cuda") 30 | 31 | @bentoml.api 32 | def txt2img(self, prompt: str = sample_prompt) -> Image: 33 | # step number to match ckpt file version 34 | num_inference_steps = 4 35 | guidance_scale = 0.0 36 | image = self.pipe( 37 | prompt=prompt, 38 | guidance_scale=guidance_scale, 39 | height=768, 40 | width=1360, 41 | num_inference_steps=num_inference_steps, 42 | max_sequence_length=256, 43 | ).images[0] 44 | return image 45 | -------------------------------------------------------------------------------- /sd3-medium/.bentoignore: -------------------------------------------------------------------------------- 1 | __pycache__/ 2 | *.py[cod] 3 | *$py.class 4 | .ipynb_checkpoints 5 | venv/ 6 | -------------------------------------------------------------------------------- /sd3-medium/README.md: -------------------------------------------------------------------------------- 1 |
2 |

Serving Stable Diffusion 3 Medium with BentoML

3 |
4 | 5 | [Stable Diffusion 3 Medium](https://huggingface.co/stabilityai/stable-diffusion-3-medium) is a Multimodal Diffusion Transformer (MMDiT) text-to-image model that features greatly improved performance in image quality, typography, complex prompt understanding, and resource-efficiency. 6 | 7 | This is a BentoML example project, demonstrating how to build an image generation inference API server, using the Stable Diffusion 3 Medium model. See [here](https://docs.bentoml.com/en/latest/examples/overview.html) for a full list of BentoML example projects. 8 | 9 | ## Prerequisites 10 | 11 | - Accept the conditions to gain access to [Stable Diffusion 3 Medium on Hugging Face](https://huggingface.co/stabilityai/stable-diffusion-3-medium). 12 | - To run the Service locally, we recommend you use an Nvidia GPU with at least 20G VRAM. 13 | 14 | ## Install dependencies 15 | 16 | ```bash 17 | git clone https://github.com/bentoml/BentoDiffusion.git 18 | cd BentoDiffusion/sd3-medium 19 | 20 | # Recommend Python 3.11 21 | pip install -r requirements.txt 22 | 23 | export HF_TOKEN= 24 | ``` 25 | 26 | ## Run the BentoML Service 27 | 28 | We have defined a BentoML Service in `service.py`. Run `bentoml serve` in your project directory to start the Service. 29 | 30 | ```bash 31 | $ bentoml serve 32 | 33 | 2024-01-18T18:31:49+0800 [INFO] [cli] Starting production HTTP BentoServer from "service:SD3Medium" listening on http://localhost:3000 (Press CTRL+C to quit) 34 | Loading pipeline components...: 100% 35 | ``` 36 | 37 | The server is now active at [http://localhost:3000](http://localhost:3000/). You can interact with it using the Swagger UI or in other different ways. 38 | 39 | CURL 40 | 41 | ```bash 42 | curl -X 'POST' \ 43 | 'http://localhost:3000/txt2img' \ 44 | -H 'accept: image/*' \ 45 | -H 'Content-Type: application/json' \ 46 | -d '{ 47 | "prompt": "A cat holding a sign that says hello world", 48 | "num_inference_steps": 28, 49 | "guidance_scale": 7.0 50 | }' 51 | ``` 52 | 53 | Python client 54 | 55 | ```python 56 | import bentoml 57 | 58 | with bentoml.SyncHTTPClient("http://localhost:3000") as client: 59 | result = client.txt2img( 60 | prompt="A cat holding a sign that says hello world", 61 | num_inference_steps=28, 62 | guidance_scale=7.0 63 | ) 64 | ``` 65 | 66 | ## Deploy to BentoCloud 67 | 68 | After the Service is ready, you can deploy the application to BentoCloud for better management and scalability. [Sign up](https://www.bentoml.com/) if you haven't got a BentoCloud account. 69 | 70 | Make sure you have [logged in to BentoCloud](https://docs.bentoml.com/en/latest/scale-with-bentocloud/manage-api-tokens.html). 71 | 72 | ```bash 73 | bentoml cloud login 74 | ``` 75 | 76 | Create a BentoCloud secret to store the required environment variable and reference it for deployment. 77 | 78 | ```bash 79 | bentoml secret create huggingface HF_TOKEN=$HF_TOKEN 80 | 81 | bentoml deploy --secret huggingface 82 | ``` 83 | 84 | Once the application is up and running on BentoCloud, you can access it via the exposed URL. 85 | 86 | **Note**: For custom deployment in your own infrastructure, use [BentoML to generate an OCI-compliant image](https://docs.bentoml.com/en/latest/get-started/packaging-for-deployment.html). 87 | -------------------------------------------------------------------------------- /sd3-medium/bentofile.yaml: -------------------------------------------------------------------------------- 1 | service: "service:SD3Medium" 2 | labels: 3 | owner: bentoml-team 4 | project: gallery 5 | include: 6 | - "*.py" 7 | python: 8 | requirements_txt: "./requirements.txt" 9 | lock_packages: false 10 | docker: 11 | python_version: "3.11" 12 | envs: 13 | - name: HF_TOKEN 14 | -------------------------------------------------------------------------------- /sd3-medium/requirements.txt: -------------------------------------------------------------------------------- 1 | accelerate==0.31.0 2 | bentoml==1.4.11 3 | diffusers==0.29.0 4 | peft==0.11.1 5 | pillow==10.3.0 6 | protobuf==5.27.1 7 | sentencepiece==0.2.0 8 | torch==2.6.0 9 | transformers==4.48.0 10 | -------------------------------------------------------------------------------- /sd3-medium/service.py: -------------------------------------------------------------------------------- 1 | import typing as t 2 | import bentoml 3 | from PIL.Image import Image 4 | from annotated_types import Le, Ge 5 | from typing_extensions import Annotated 6 | 7 | 8 | MODEL_ID = "stabilityai/stable-diffusion-3-medium-diffusers" 9 | 10 | sample_prompt = "A cat holding a sign that says hello world" 11 | 12 | @bentoml.service( 13 | traffic={"timeout": 300}, 14 | workers=1, 15 | resources={ 16 | "gpu": 1, 17 | "gpu_type": "nvidia-l4", 18 | }, 19 | ) 20 | class SD3Medium: 21 | def __init__(self) -> None: 22 | import torch 23 | from diffusers import StableDiffusion3Pipeline 24 | 25 | self.pipe = StableDiffusion3Pipeline.from_pretrained( 26 | MODEL_ID, 27 | torch_dtype=torch.float16, 28 | ) 29 | self.pipe.to(device="cuda") 30 | 31 | @bentoml.api 32 | def txt2img( 33 | self, 34 | prompt: str = sample_prompt, 35 | negative_prompt: t.Optional[str] = None, 36 | num_inference_steps: Annotated[int, Ge(1), Le(50)] = 28, 37 | guidance_scale: float = 7.0, 38 | ) -> Image: 39 | image = self.pipe( 40 | prompt=prompt, 41 | negative_prompt=negative_prompt, 42 | num_inference_steps=num_inference_steps, 43 | guidance_scale=guidance_scale, 44 | ).images[0] 45 | return image 46 | -------------------------------------------------------------------------------- /sd3.5-large-turbo/.bentoignore: -------------------------------------------------------------------------------- 1 | __pycache__/ 2 | *.py[cod] 3 | *$py.class 4 | .ipynb_checkpoints 5 | venv/ 6 | -------------------------------------------------------------------------------- /sd3.5-large-turbo/README.md: -------------------------------------------------------------------------------- 1 |
2 |

Serving Stable Diffusion 3.5 Large Turbo with BentoML

3 |
4 | 5 | [Stable Diffusion 3.5 Large Turbo](https://huggingface.co/stabilityai/stable-diffusion-3.5-large-turbo) is a Multimodal Diffusion Transformer (MMDiT) text-to-image model with Adversarial Diffusion Distillation (ADD) that features improved performance in image quality, typography, complex prompt understanding, and resource-efficiency, with a focus on fewer inference steps. 6 | 7 | This is a BentoML example project, demonstrating how to build an image generation inference API server, using the Stable Diffusion 3.5 Large Turbo model. See [here](https://docs.bentoml.com/en/latest/examples/overview.html) for a full list of BentoML example projects. 8 | 9 | ## Prerequisites 10 | 11 | - Accept the conditions to gain access to [Stable Diffusion 3.5 Large Turbo on Hugging Face](https://huggingface.co/stabilityai/stable-diffusion-3.5-large-turbo). 12 | - To run the Service locally, you need an Nvidia GPU with at least 32G VRAM. 13 | 14 | ## Install dependencies 15 | 16 | ```bash 17 | git clone https://github.com/bentoml/BentoDiffusion.git 18 | cd BentoDiffusion/sd3.5-large-turbo 19 | pip install -r requirements.txt 20 | 21 | export HF_TOKEN= 22 | ``` 23 | 24 | ## Run the BentoML Service 25 | 26 | We have defined a BentoML Service in `service.py`. Run `bentoml serve` in your project directory to start the Service. 27 | 28 | ```bash 29 | $ bentoml serve 30 | 31 | 2024-01-18T18:31:49+0800 [INFO] [cli] Starting production HTTP BentoServer from "service:SD3.5LargeTurbo" listening on http://localhost:3000 (Press CTRL+C to quit) 32 | Loading pipeline components...: 100% 33 | ``` 34 | 35 | The server is now active at [http://localhost:3000](http://localhost:3000/). You can interact with it using the Swagger UI or in other different ways. 36 | 37 | CURL 38 | 39 | ```bash 40 | curl -X 'POST' \ 41 | 'http://localhost:3000/txt2img' \ 42 | -H 'accept: image/*' \ 43 | -H 'Content-Type: application/json' \ 44 | -d '{ 45 | "prompt": "A cat holding a sign that says hello world", 46 | "num_inference_steps": 4 47 | }' 48 | ``` 49 | 50 | Python client 51 | 52 | ```python 53 | import bentoml 54 | 55 | with bentoml.SyncHTTPClient("http://localhost:3000") as client: 56 | result = client.txt2img( 57 | prompt="A cat holding a sign that says hello world", 58 | num_inference_steps=4 59 | ) 60 | ``` 61 | 62 | ## Deploy to BentoCloud 63 | 64 | After the Service is ready, you can deploy the application to BentoCloud for better management and scalability. [Sign up](https://www.bentoml.com/) if you haven't got a BentoCloud account. 65 | 66 | Make sure you have [logged in to BentoCloud](https://docs.bentoml.com/en/latest/scale-with-bentocloud/manage-api-tokens.html). 67 | 68 | ```bash 69 | bentoml cloud login 70 | ``` 71 | 72 | Create a BentoCloud secret to store the required environment variable and reference it for deployment. 73 | 74 | ```bash 75 | bentoml secret create huggingface HF_TOKEN=$HF_TOKEN 76 | 77 | bentoml deploy --secret huggingface 78 | ``` 79 | 80 | Once the application is up and running on BentoCloud, you can access it via the exposed URL. 81 | 82 | **Note**: For custom deployment in your own infrastructure, use [BentoML to generate an OCI-compliant image](https://docs.bentoml.com/en/latest/get-started/packaging-for-deployment.html). 83 | -------------------------------------------------------------------------------- /sd3.5-large-turbo/bentofile.yaml: -------------------------------------------------------------------------------- 1 | service: "service:SD35LargeTurbo" 2 | labels: 3 | owner: bentoml-team 4 | project: gallery 5 | include: 6 | - "*.py" 7 | python: 8 | requirements_txt: "./requirements.txt" 9 | lock_packages: false 10 | envs: 11 | - name: HF_TOKEN 12 | -------------------------------------------------------------------------------- /sd3.5-large-turbo/requirements.txt: -------------------------------------------------------------------------------- 1 | accelerate==1.0.1 2 | bentoml==1.4.11 3 | diffusers==0.31.0 4 | pillow==11.0.0 5 | protobuf==5.28.3 6 | sentencepiece==0.2.0 7 | torch==2.6.0 8 | transformers==4.48.0 9 | -------------------------------------------------------------------------------- /sd3.5-large-turbo/service.py: -------------------------------------------------------------------------------- 1 | import typing as t 2 | import bentoml 3 | from PIL.Image import Image 4 | from annotated_types import Le, Ge 5 | from typing_extensions import Annotated 6 | 7 | 8 | MODEL_ID = "stabilityai/stable-diffusion-3.5-large-turbo" 9 | 10 | sample_prompt = "A cat holding a sign that says hello world" 11 | 12 | @bentoml.service( 13 | traffic={"timeout": 300}, 14 | workers=1, 15 | resources={ 16 | "gpu": 1, 17 | "gpu_type": "nvidia-tesla-a100", 18 | }, 19 | ) 20 | class SD35LargeTurbo: 21 | def __init__(self) -> None: 22 | import torch 23 | from diffusers import StableDiffusion3Pipeline 24 | 25 | self.pipe = StableDiffusion3Pipeline.from_pretrained( 26 | MODEL_ID, 27 | torch_dtype=torch.bfloat16, 28 | ) 29 | self.pipe.to(device="cuda") 30 | 31 | @bentoml.api 32 | def txt2img( 33 | self, 34 | prompt: str = sample_prompt, 35 | negative_prompt: t.Optional[str] = None, 36 | num_inference_steps: Annotated[int, Ge(1), Le(10)] = 4, 37 | ) -> Image: 38 | image = self.pipe( 39 | prompt=prompt, 40 | negative_prompt=negative_prompt, 41 | num_inference_steps=num_inference_steps, 42 | guidance_scale=0.0, 43 | ).images[0] 44 | return image 45 | -------------------------------------------------------------------------------- /sd3.5-large/.bentoignore: -------------------------------------------------------------------------------- 1 | __pycache__/ 2 | *.py[cod] 3 | *$py.class 4 | .ipynb_checkpoints 5 | venv/ 6 | -------------------------------------------------------------------------------- /sd3.5-large/README.md: -------------------------------------------------------------------------------- 1 |
2 |

Serving Stable Diffusion 3.5 Large with BentoML

3 |
4 | 5 | [Stable Diffusion 3.5 Large](https://huggingface.co/stabilityai/stable-diffusion-3.5-large) is a Multimodal Diffusion Transformer (MMDiT) text-to-image model that features improved performance in image quality, typography, complex prompt understanding, and resource-efficiency. 6 | 7 | This is a BentoML example project, demonstrating how to build an image generation inference API server, using the Stable Diffusion 3.5 Large model. See [here](https://docs.bentoml.com/en/latest/examples/overview.html) for a full list of BentoML example projects. 8 | 9 | ## Prerequisites 10 | 11 | - Accept the conditions to gain access to [Stable Diffusion 3.5 Large on Hugging Face](https://huggingface.co/stabilityai/stable-diffusion-3.5-large). 12 | - To run the Service locally, you need an Nvidia GPU with at least 20G VRAM. 13 | 14 | ## Install dependencies 15 | 16 | ```bash 17 | git clone https://github.com/bentoml/BentoDiffusion.git 18 | cd BentoDiffusion/sd3.5-large 19 | pip install -r requirements.txt 20 | 21 | export HF_TOKEN= 22 | ``` 23 | 24 | ## Run the BentoML Service 25 | 26 | We have defined a BentoML Service in `service.py`. Run `bentoml serve` in your project directory to start the Service. 27 | 28 | ```bash 29 | $ bentoml serve 30 | 31 | 2024-01-18T18:31:49+0800 [INFO] [cli] Starting production HTTP BentoServer from "service:SD35Large" listening on http://localhost:3000 (Press CTRL+C to quit) 32 | Loading pipeline components...: 100% 33 | ``` 34 | 35 | The server is now active at [http://localhost:3000](http://localhost:3000/). You can interact with it using the Swagger UI or in other different ways. 36 | 37 | CURL 38 | 39 | ```bash 40 | curl -X 'POST' \ 41 | 'http://localhost:3000/txt2img' \ 42 | -H 'accept: image/*' \ 43 | -H 'Content-Type: application/json' \ 44 | -d '{ 45 | "prompt": "A cat holding a sign that says hello world", 46 | "num_inference_steps": 40, 47 | "guidance_scale": 4.5 48 | }' 49 | ``` 50 | 51 | Python client 52 | 53 | ```python 54 | import bentoml 55 | 56 | with bentoml.SyncHTTPClient("http://localhost:3000") as client: 57 | result = client.txt2img( 58 | prompt="A cat holding a sign that says hello world", 59 | num_inference_steps=40, 60 | guidance_scale=4.5 61 | ) 62 | ``` 63 | 64 | ## Deploy to BentoCloud 65 | 66 | After the Service is ready, you can deploy the application to BentoCloud for better management and scalability. [Sign up](https://www.bentoml.com/) if you haven't got a BentoCloud account. 67 | 68 | Make sure you have [logged in to BentoCloud](https://docs.bentoml.com/en/latest/scale-with-bentocloud/manage-api-tokens.html). 69 | 70 | ```bash 71 | bentoml cloud login 72 | ``` 73 | 74 | Create a BentoCloud secret to store the required environment variable and reference it for deployment. 75 | 76 | ```bash 77 | bentoml secret create huggingface HF_TOKEN=$HF_TOKEN 78 | 79 | bentoml deploy --secret huggingface 80 | ``` 81 | 82 | Once the application is up and running on BentoCloud, you can access it via the exposed URL. 83 | 84 | **Note**: For custom deployment in your own infrastructure, use [BentoML to generate an OCI-compliant image](https://docs.bentoml.com/en/latest/get-started/packaging-for-deployment.html). 85 | -------------------------------------------------------------------------------- /sd3.5-large/bentofile.yaml: -------------------------------------------------------------------------------- 1 | service: "service:SD35Large" 2 | labels: 3 | owner: bentoml-team 4 | project: gallery 5 | include: 6 | - "*.py" 7 | python: 8 | requirements_txt: "./requirements.txt" 9 | lock_packages: false 10 | envs: 11 | - name: HF_TOKEN 12 | -------------------------------------------------------------------------------- /sd3.5-large/requirements.txt: -------------------------------------------------------------------------------- 1 | accelerate==1.0.1 2 | bentoml==1.4.11 3 | diffusers==0.31.0 4 | pillow==11.0.0 5 | protobuf==5.28.3 6 | sentencepiece==0.2.0 7 | torch==2.6.0 8 | transformers==4.48.0 9 | -------------------------------------------------------------------------------- /sd3.5-large/service.py: -------------------------------------------------------------------------------- 1 | import typing as t 2 | import bentoml 3 | from PIL.Image import Image 4 | from annotated_types import Le, Ge 5 | from typing_extensions import Annotated 6 | 7 | 8 | MODEL_ID = "stabilityai/stable-diffusion-3.5-large" 9 | 10 | sample_prompt = "A cat holding a sign that says hello world" 11 | 12 | @bentoml.service( 13 | traffic={"timeout": 300}, 14 | workers=1, 15 | resources={ 16 | "gpu": 1, 17 | "gpu_type": "nvidia-tesla-a100", 18 | }, 19 | ) 20 | class SD35Large: 21 | def __init__(self) -> None: 22 | import torch 23 | from diffusers import StableDiffusion3Pipeline 24 | 25 | self.pipe = StableDiffusion3Pipeline.from_pretrained( 26 | MODEL_ID, 27 | torch_dtype=torch.bfloat16, 28 | ) 29 | self.pipe.to(device="cuda") 30 | 31 | @bentoml.api 32 | def txt2img( 33 | self, 34 | prompt: str = sample_prompt, 35 | negative_prompt: t.Optional[str] = None, 36 | num_inference_steps: Annotated[int, Ge(1), Le(50)] = 40, 37 | guidance_scale: float = 4.5, 38 | ) -> Image: 39 | image = self.pipe( 40 | prompt=prompt, 41 | negative_prompt=negative_prompt, 42 | num_inference_steps=num_inference_steps, 43 | guidance_scale=guidance_scale, 44 | ).images[0] 45 | return image 46 | -------------------------------------------------------------------------------- /sdxl-lightning/.bentoignore: -------------------------------------------------------------------------------- 1 | __pycache__/ 2 | *.py[cod] 3 | *$py.class 4 | .ipynb_checkpoints 5 | venv/ 6 | -------------------------------------------------------------------------------- /sdxl-lightning/README.md: -------------------------------------------------------------------------------- 1 |
2 |

Serving SDXL Lightning with BentoML

3 |
4 | 5 | This is a BentoML example project, demonstrating how to build an image generation inference API server using the [SDXL-Lightning model](https://huggingface.co/ByteDance/SDXL-Lightning), a lightning-fast text-to-image generation model that is able to generate high-quality 1024px images in a few steps. 6 | 7 | See [here](https://docs.bentoml.com/en/latest/examples/overview.html) for a full list of BentoML example projects. 8 | 9 | ## Prerequisites 10 | 11 | To run the Service locally, we recommend you use an Nvidia GPU with at least 16G VRAM. 12 | 13 | ## Install dependencies 14 | 15 | ```bash 16 | git clone https://github.com/bentoml/BentoDiffusion.git 17 | cd BentoDiffusion/sdxl-lightning 18 | 19 | # Recommend Python 3.11 20 | pip install -r requirements.txt 21 | 22 | export HF_TOKEN= 23 | ``` 24 | 25 | ## Run the BentoML Service 26 | 27 | We have defined a BentoML Service in `service.py`. Run `bentoml serve` in your project directory to start the Service. 28 | 29 | ```bash 30 | $ bentoml serve 31 | 32 | 2024-01-18T18:31:49+0800 [INFO] [cli] Starting production HTTP BentoServer from "service:SDXLLightning" listening on http://localhost:3000 (Press CTRL+C to quit) 33 | Loading pipeline components...: 100% 34 | ``` 35 | 36 | The server is now active at [http://localhost:3000](http://localhost:3000/). You can interact with it using the Swagger UI or in other different ways. 37 | 38 | CURL 39 | 40 | ```bash 41 | curl -X 'POST' \ 42 | 'http://localhost:3000/txt2img' \ 43 | -H 'accept: image/*' \ 44 | -H 'Content-Type: application/json' \ 45 | -d '{ 46 | "prompt": "A cinematic shot of a baby racoon wearing an intricate italian priest robe.", 47 | "num_inference_steps": 1, 48 | "guidance_scale": 0 49 | }' 50 | ``` 51 | 52 | BentoML client 53 | 54 | ```python 55 | import bentoml 56 | 57 | with bentoml.SyncHTTPClient("http://localhost:3000") as client: 58 | result = client.txt2img( 59 | prompt="A cinematic shot of a baby racoon wearing an intricate italian priest robe.", 60 | num_inference_steps=1, 61 | guidance_scale=0.0 62 | ) 63 | ``` 64 | 65 | ## Deploy to BentoCloud 66 | 67 | After the Service is ready, you can deploy the application to BentoCloud for better management and scalability. [Sign up](https://www.bentoml.com/) if you haven't got a BentoCloud account. 68 | 69 | Make sure you have [logged in to BentoCloud](https://docs.bentoml.com/en/latest/scale-with-bentocloud/manage-api-tokens.html). 70 | 71 | ```bash 72 | bentoml cloud login 73 | ``` 74 | 75 | Deploy it to BentoCloud. 76 | 77 | ```bash 78 | bentoml deploy 79 | ``` 80 | 81 | Once the application is up and running on BentoCloud, you can access it via the exposed URL. 82 | 83 | **Note**: For custom deployment in your own infrastructure, use [BentoML to generate an OCI-compliant image](https://docs.bentoml.com/en/latest/get-started/packaging-for-deployment.html). 84 | -------------------------------------------------------------------------------- /sdxl-lightning/bentofile.yaml: -------------------------------------------------------------------------------- 1 | service: "service:SDXLLightning" 2 | labels: 3 | owner: bentoml-team 4 | project: gallery 5 | include: 6 | - "*.py" 7 | python: 8 | requirements_txt: "./requirements.txt" 9 | docker: 10 | python_version: "3.11" 11 | -------------------------------------------------------------------------------- /sdxl-lightning/requirements.txt: -------------------------------------------------------------------------------- 1 | accelerate==0.31.0 2 | bentoml==1.4.11 3 | diffusers==0.29.0 4 | pillow==10.3.0 5 | torch==2.6.0 6 | transformers==4.48.0 7 | -------------------------------------------------------------------------------- /sdxl-lightning/service.py: -------------------------------------------------------------------------------- 1 | import bentoml 2 | from PIL.Image import Image 3 | from annotated_types import Le, Ge 4 | from typing_extensions import Annotated 5 | 6 | BASE_MODEL_ID = "stabilityai/stable-diffusion-xl-base-1.0" 7 | REPO = "ByteDance/SDXL-Lightning" 8 | CKPT = "sdxl_lightning_4step_unet.safetensors" # Use the correct ckpt for your step setting! 9 | 10 | sample_prompt = "A girl smiling" 11 | 12 | @bentoml.service( 13 | traffic={"timeout": 300}, 14 | workers=1, 15 | resources={ 16 | "gpu": 1, 17 | "gpu_type": "nvidia-l4", 18 | }, 19 | ) 20 | class SDXLLightning: 21 | def __init__(self) -> None: 22 | import torch 23 | from diffusers import ( 24 | StableDiffusionXLPipeline, 25 | UNet2DConditionModel, 26 | EulerDiscreteScheduler 27 | ) 28 | from huggingface_hub import hf_hub_download 29 | from safetensors.torch import load_file 30 | 31 | self.unet = UNet2DConditionModel.from_config( 32 | BASE_MODEL_ID, subfolder="unet" 33 | ).to("cuda", torch.float16) 34 | 35 | self.unet.load_state_dict( 36 | load_file(hf_hub_download(REPO, CKPT), 37 | device="cuda") 38 | ) 39 | 40 | self.pipe = StableDiffusionXLPipeline.from_pretrained( 41 | BASE_MODEL_ID, 42 | unet=self.unet, 43 | torch_dtype=torch.float16, 44 | variant="fp16" 45 | ).to("cuda") 46 | 47 | self.pipe.scheduler = EulerDiscreteScheduler.from_config( 48 | self.pipe.scheduler.config, timestep_spacing="trailing" 49 | ) 50 | 51 | 52 | @bentoml.api 53 | def txt2img(self, prompt: str = sample_prompt) -> Image: 54 | # step number to match ckpt file version 55 | num_inference_steps = 4 56 | guidance_scale = 0.0 57 | image = self.pipe( 58 | prompt=prompt, 59 | num_inference_steps=num_inference_steps, 60 | guidance_scale=guidance_scale, 61 | ).images[0] 62 | return image 63 | -------------------------------------------------------------------------------- /sdxl-turbo/.bentoignore: -------------------------------------------------------------------------------- 1 | __pycache__/ 2 | *.py[cod] 3 | *$py.class 4 | .ipynb_checkpoints 5 | venv/ 6 | -------------------------------------------------------------------------------- /sdxl-turbo/README.md: -------------------------------------------------------------------------------- 1 |
2 |

Serving SDXL Turbo with BentoML

3 |
4 | 5 | [Stable Diffusion XL Turbo](https://huggingface.co/stabilityai/sdxl-turbo) is a real-time text-to-image generation model utilizing a novel distillation technique called Adversarial Diffusion Distillation (ADD). This technology enables SDXL Turbo to generate images in a single step, significantly enhancing performance and reducing computational requirements without sacrificing image quality. 6 | 7 | This is a BentoML example project, demonstrating how to build an image generation inference API server, using the SDXL Turbo model. See [here](https://docs.bentoml.com/en/latest/examples/overview.html) for a full list of BentoML example projects. 8 | 9 | ## Prerequisites 10 | 11 | To run the Service locally, we recommend you use an Nvidia GPU with at least 12G VRAM. 12 | 13 | ## Install dependencies 14 | 15 | ```bash 16 | git clone https://github.com/bentoml/BentoDiffusion.git 17 | cd BentoDiffusion/sdxl-turbo 18 | 19 | # Recommend Python 3.11 20 | pip install -r requirements.txt 21 | ``` 22 | 23 | ## Run the BentoML Service 24 | 25 | We have defined a BentoML Service in `service.py`. Run `bentoml serve` in your project directory to start the Service. 26 | 27 | ```bash 28 | $ bentoml serve 29 | 30 | 2024-01-18T18:31:49+0800 [INFO] [cli] Starting production HTTP BentoServer from "service:SDXLTurboService" listening on http://localhost:3000 (Press CTRL+C to quit) 31 | Loading pipeline components...: 100% 32 | ``` 33 | 34 | The server is now active at [http://localhost:3000](http://localhost:3000/). You can interact with it using the Swagger UI or in other different ways. 35 | 36 | CURL 37 | 38 | ```bash 39 | curl -X 'POST' \ 40 | 'http://localhost:3000/txt2img' \ 41 | -H 'accept: image/*' \ 42 | -H 'Content-Type: application/json' \ 43 | -d '{ 44 | "prompt": "A cinematic shot of a baby racoon wearing an intricate italian priest robe.", 45 | "num_inference_steps": 1, 46 | "guidance_scale": 0 47 | }' 48 | ``` 49 | 50 | Python client 51 | 52 | ```python 53 | import bentoml 54 | 55 | with bentoml.SyncHTTPClient("http://localhost:3000") as client: 56 | result = client.txt2img( 57 | prompt="A cinematic shot of a baby racoon wearing an intricate italian priest robe.", 58 | num_inference_steps=1, 59 | guidance_scale=0.0 60 | ) 61 | ``` 62 | 63 | For detailed explanations of the Service code, see [Stable Diffusion XL Turbo](https://docs.bentoml.com/en/latest/use-cases/diffusion-models/sdxl-turbo.html). 64 | 65 | ## Deploy to BentoCloud 66 | 67 | After the Service is ready, you can deploy the application to BentoCloud for better management and scalability. [Sign up](https://www.bentoml.com/) if you haven't got a BentoCloud account. 68 | 69 | Make sure you have [logged in to BentoCloud](https://docs.bentoml.com/en/latest/scale-with-bentocloud/manage-api-tokens.html). 70 | 71 | ```bash 72 | bentoml cloud login 73 | ``` 74 | 75 | Deploy it to BentoCloud. 76 | 77 | ```bash 78 | bentoml deploy 79 | ``` 80 | 81 | Once the application is up and running on BentoCloud, you can access it via the exposed URL. 82 | 83 | **Note**: For custom deployment in your own infrastructure, use [BentoML to generate an OCI-compliant image](https://docs.bentoml.com/en/latest/get-started/packaging-for-deployment.html). 84 | -------------------------------------------------------------------------------- /sdxl-turbo/requirements.txt: -------------------------------------------------------------------------------- 1 | accelerate==0.31.0 2 | bentoml==1.4.11 3 | diffusers==0.29.0 4 | pillow==10.3.0 5 | torch==2.6.0 6 | transformers==4.48.0 7 | -------------------------------------------------------------------------------- /sdxl-turbo/service.py: -------------------------------------------------------------------------------- 1 | import bentoml 2 | from PIL.Image import Image 3 | from annotated_types import Le, Ge 4 | from typing_extensions import Annotated 5 | 6 | 7 | MODEL_ID = "stabilityai/sdxl-turbo" 8 | 9 | sample_prompt = "A cinematic shot of a baby racoon wearing an intricate italian priest robe." 10 | 11 | my_image = bentoml.images.PythonImage(python_version="3.11") \ 12 | .requirements_file("requirements.txt") 13 | 14 | @bentoml.service( 15 | image=my_image, 16 | traffic={"timeout": 300}, 17 | workers=1, 18 | labels={'owner': 'bentoml-team', 'project': 'gallery'}, 19 | resources={ 20 | "gpu": 1, 21 | "gpu_type": "nvidia-l4", 22 | }, 23 | ) 24 | class SDXLTurbo: 25 | model_path = bentoml.models.HuggingFaceModel(MODEL_ID) 26 | 27 | def __init__(self) -> None: 28 | from diffusers import AutoPipelineForText2Image 29 | import torch 30 | 31 | self.pipe = AutoPipelineForText2Image.from_pretrained( 32 | self.model_path, 33 | torch_dtype=torch.float16, 34 | variant="fp16", 35 | ) 36 | self.pipe.to(device="cuda") 37 | 38 | @bentoml.api 39 | def txt2img( 40 | self, 41 | prompt: str = sample_prompt, 42 | num_inference_steps: Annotated[int, Ge(1), Le(10)] = 1, 43 | guidance_scale: float = 0.0, 44 | ) -> Image: 45 | image = self.pipe( 46 | prompt=prompt, 47 | num_inference_steps=num_inference_steps, 48 | guidance_scale=guidance_scale, 49 | ).images[0] 50 | return image 51 | --------------------------------------------------------------------------------