|
| 1 | +<!--Copyright 2024 The HuggingFace Team, The Black Forest Team. All rights reserved. |
| 2 | +
|
| 3 | +Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with |
| 4 | +the License. You may obtain a copy of the License at |
| 5 | +
|
| 6 | +http://www.apache.org/licenses/LICENSE-2.0 |
| 7 | +
|
| 8 | +Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on |
| 9 | +an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the |
| 10 | +specific language governing permissions and limitations under the License. |
| 11 | +--> |
| 12 | + |
| 13 | +# FluxControlInpaint |
| 14 | + |
| 15 | +FluxControlInpaintPipeline is an implementation of Inpainting for Flux.1 Depth/Canny models. It is a pipeline that allows you to inpaint images using the Flux.1 Depth/Canny models. The pipeline takes an image and a mask as input and returns the inpainted image. |
| 16 | + |
| 17 | +FLUX.1 Depth and Canny [dev] is a 12 billion parameter rectified flow transformer capable of generating an image based on a text description while following the structure of a given input image. **This is not a ControlNet model**. |
| 18 | + |
| 19 | +| Control type | Developer | Link | |
| 20 | +| -------- | ---------- | ---- | |
| 21 | +| Depth | [Black Forest Labs](https://huggingface.co/black-forest-labs) | [Link](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev) | |
| 22 | +| Canny | [Black Forest Labs](https://huggingface.co/black-forest-labs) | [Link](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev) | |
| 23 | + |
| 24 | + |
| 25 | +<Tip> |
| 26 | + |
| 27 | +Flux can be quite expensive to run on consumer hardware devices. However, you can perform a suite of optimizations to run it faster and in a more memory-friendly manner. Check out [this section](https://huggingface.co/blog/sd3#memory-optimizations-for-sd3) for more details. Additionally, Flux can benefit from quantization for memory efficiency with a trade-off in inference latency. Refer to [this blog post](https://huggingface.co/blog/quanto-diffusers) to learn more. For an exhaustive list of resources, check out [this gist](https://gist.github.com/sayakpaul/b664605caf0aa3bf8585ab109dd5ac9c). |
| 28 | + |
| 29 | +</Tip> |
| 30 | + |
| 31 | +```python |
| 32 | +import torch |
| 33 | +from diffusers import FluxControlInpaintPipeline |
| 34 | +from diffusers.models.transformers import FluxTransformer2DModel |
| 35 | +from transformers import T5EncoderModel |
| 36 | +from diffusers.utils import load_image, make_image_grid |
| 37 | +from image_gen_aux import DepthPreprocessor # https://github.com/huggingface/image_gen_aux |
| 38 | +from PIL import Image |
| 39 | +import numpy as np |
| 40 | + |
| 41 | +pipe = FluxControlInpaintPipeline.from_pretrained( |
| 42 | + "black-forest-labs/FLUX.1-Depth-dev", |
| 43 | + torch_dtype=torch.bfloat16, |
| 44 | +) |
| 45 | +# use following lines if you have GPU constraints |
| 46 | +# --------------------------------------------------------------- |
| 47 | +transformer = FluxTransformer2DModel.from_pretrained( |
| 48 | + "sayakpaul/FLUX.1-Depth-dev-nf4", subfolder="transformer", torch_dtype=torch.bfloat16 |
| 49 | +) |
| 50 | +text_encoder_2 = T5EncoderModel.from_pretrained( |
| 51 | + "sayakpaul/FLUX.1-Depth-dev-nf4", subfolder="text_encoder_2", torch_dtype=torch.bfloat16 |
| 52 | +) |
| 53 | +pipe.transformer = transformer |
| 54 | +pipe.text_encoder_2 = text_encoder_2 |
| 55 | +pipe.enable_model_cpu_offload() |
| 56 | +# --------------------------------------------------------------- |
| 57 | +pipe.to("cuda") |
| 58 | + |
| 59 | +prompt = "a blue robot singing opera with human-like expressions" |
| 60 | +image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/robot.png") |
| 61 | + |
| 62 | +head_mask = np.zeros_like(image) |
| 63 | +head_mask[65:580,300:642] = 255 |
| 64 | +mask_image = Image.fromarray(head_mask) |
| 65 | + |
| 66 | +processor = DepthPreprocessor.from_pretrained("LiheYoung/depth-anything-large-hf") |
| 67 | +control_image = processor(image)[0].convert("RGB") |
| 68 | + |
| 69 | +output = pipe( |
| 70 | + prompt=prompt, |
| 71 | + image=image, |
| 72 | + control_image=control_image, |
| 73 | + mask_image=mask_image, |
| 74 | + num_inference_steps=30, |
| 75 | + strength=0.9, |
| 76 | + guidance_scale=10.0, |
| 77 | + generator=torch.Generator().manual_seed(42), |
| 78 | +).images[0] |
| 79 | +make_image_grid([image, control_image, mask_image, output.resize(image.size)], rows=1, cols=4).save("output.png") |
| 80 | +``` |
| 81 | + |
| 82 | +## FluxControlInpaintPipeline |
| 83 | +[[autodoc]] FluxControlInpaintPipeline |
| 84 | + - all |
| 85 | + - __call__ |
| 86 | + |
| 87 | + |
| 88 | +## FluxPipelineOutput |
| 89 | +[[autodoc]] pipelines.flux.pipeline_output.FluxPipelineOutput |
0 commit comments