metadata

license: other
license_name: flux-1-dev-non-commercial-license
license_link: https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md
language:
  - en
library_name: diffusers
pipeline_tag: text-to-image
tags:
  - Text-to-Image
  - ControlNet
  - Diffusers
  - Flux.1-dev
  - image-generation
  - Stable Diffusion
base_model: black-forest-labs/FLUX.1-dev

FLUX.1-dev-ControlNet-Depth

This repository contains a Depth ControlNet for FLUX.1-dev model jointly trained by researchers from InstantX Team and Shakker Labs.

Model Cards

The model consists of 4 FluxTransformerBlock and 1 FluxSingleTransformerBlock.
This checkpoint is trained on both real and generated image datasets, with 16*A800 for 70K steps. The batch size 16*4=64 with resolution=1024. The learning rate is set to 5e-6. We use Depth-Anything-V2 to extract depth maps.
The recommended controlnet_conditioning_scale is 0.3-0.7.

Showcases

Inference

import torch
from diffusers.utils import load_image
from diffusers import FluxControlNetPipeline, FluxControlNetModel

controlnet_model = "black-forest-labs/FLUX.1-dev"
base_model = "Shakker-Labs/FLUX.1-dev-ControlNet-Depth"

controlnet = FluxControlNetModel.from_pretrained(controlnet_model, torch_dtype=torch.bfloat16)
pipe = FluxControlNetPipeline.from_pretrained(
    base_model, controlnet=controlnet, torch_dtype=torch.bfloat16
)
pipe.to("cuda")

control_image = load_image("https://huggingface.co/Shakker-Labs/FLUX.1-dev-ControlNet-Depth/resolve/main/assets/cond1.png")
prompt = "an old man with white hair"

image = pipe(prompt,
             control_image=control_image,
             controlnet_conditioning_scale=0.5,
             width=control_image.size[0],
             height=control_image.size[1],
             num_inference_steps=24,
             guidance_scale=3.5,
).images[0]

For multi-ControlNets support, please refer to Shakker-Labs/FLUX.1-dev-ControlNet-Union-Pro.

Resources

Acknowledgements

This project is sponsored and released by Shakker AI. All copyright reserved.