Zum Inhalt springen
ModelAufgelistetExtern
Herunterladen

Flux ControlNet Training Example

this is a debug trainning result for diffusers training script

ArtControlNetBasismodellblack-forest-labs/FLUX.1-devVersiondiffusion_pytorch_modelCBewertung

Modellquelle

Art
ControlNet
Dateigröße
5,37 GB
Basismodell
black-forest-labs/FLUX.1-dev
Version
vdiffusion_pytorch_model
Aktualisiert
6. Okt. 2026
Downloads
8
Quelle
Hugging Face

Quellenbeschreibung

this is a debug trainning result for diffusers training script

related issue https://github.com/huggingface/diffusers/issues/11181

related wandb log https://wandb.ai/wuuutiiing/flux_train_controlnet/overview

related script https://github.com/huggingface/diffusers/blob/main/examples/controlnet/train_controlnet_flux.py

test script



from diffusers import FluxControlNetModel, FluxControlNetPipeline
import torch
from PIL import Image

validation_image = Image.open("a_control_image_like_fill50k.png").convert("RGB")

control_path = "wuutiing2/flux_controlnet_training_example"
flux_path = "black-forest-labs/FLUX.1-dev"

flux_controlnet = FluxControlNetModel.from_pretrained(
    control_path, torch_dtype=torch.bfloat16
)
pipeline = FluxControlNetPipeline.from_pretrained(
    flux_path,
    controlnet=flux_controlnet,
    torch_dtype=torch.bfloat16,
        )
pipeline.to("cuda")

image = pipeline(
    prompt="aqua circle with light pink background",
    control_image=validation_image,
    num_inference_steps=28,
    controlnet_conditioning_scale=1,
    guidance_scale=3.5,
    generator=torch.manual_seed(19),
).images[0]

image.save("output.png")

Quellen

1 Quelle
Hugging FacePrimärVersion ec3bf43b44c6b6a57a95c5d706e027eb37f80472
huggingface.co/wuutiing2/flux_controlnet_training_example/tree/ec3bf43b44c6b6a57a95c5d706e027eb37f80472

Verifiziert 6. Okt.

Modellartefakte

1 Artefakt

diffusion_pytorch_model.safetensors

safetensors · 5,37 GB · SHA-256 78d5b3e93eb0…062f · Hugging Face

Herunterladen

Quellenauszüge

3 Auszüge

--- license: apache-2.0 datasets: - HighCWu/fill50k language: - en base_model: - black-forest-labs/FLUX.1-dev pipeline_tag: image-to-image tags: - controlnet - Controlnet --- this is a debug trainning result for diffusers training script related issue https://github.com/huggingface/diffusers/issues/11181 related wandb log https://wandb.ai/wuuutiiing/flux_train_controlnet/overview related script https://github.com/huggingface/diffusers/blob/main/examples/controlnet/train_controlnet_flux.py # test script ``` from diffusers import FluxControlNetModel, FluxControlNetPipeline import torch from PIL import Image validation_image = Image.open("a_control_image_like_fill50k.png").convert("RGB") control_path = "wuutiing2/flux_controlnet_training_example" flux_path = "black-forest-labs/FLUX.1-dev" flux_controlnet = FluxControlNetModel.from_pretrained( control_path, torch_dtype=torch.bfloat16 ) pipeline = FluxControlNetPipeline.from_pretrained( flux_path, controlnet=flux_controlnet, torch_dtype=torch.bfloat16, ) pipeline.to("cuda") image = pipeline( prompt="aqua circle with light pink background", control_image=validation_image, num_inference_steps=28, controlnet_conditioning_scale=1, guidance_scale=3.5, generator=torch.manual_seed(19), ).images[0] image.save("output.png") ```

Source context: 8 downloads · 0 likes · Pipeline image-to-image · Library diffusers · Repo wuutiing2/flux_controlnet_training_example

Source context: 1 downloads · 0 likes · Pipeline image-to-image · Library diffusers · Repo wuutiing2/flux_controlnet_training_example