How to add flux1-fill-dev-fp8.safetensors

[SlimRG]

Describe the bug

Hi!
How to use flux1-fill-dev-fp8.safetensors in diffusers?

Now I have code:

def init_pipeline(device: str):
    logger.info(f"Loading FLUX Inpaint Pipeline (Fill‑dev) on {device}")
    pipe = FluxFillPipeline.from_pretrained(
        "black-forest-labs/FLUX.1-Fill-dev",
        torch_dtype=torch.bfloat16,
        trust_remote_code=True
    ).to(device)
    logger.info("Pipeline loaded successfully")
    return pipe

Another try:

 transformer = FluxTransformer2DModel.from_single_file(
        "https://huggingface.co/YarvixPA/FLUX.1-Fill-dev-gguf/blob/main/flux1-fill-dev-Q4_0.gguf",
        quantization_config=GGUFQuantizationConfig(compute_dtype=torch.bfloat16),
        torch_dtype=torch.bfloat16
    )

    pipe = FluxFillPipeline.from_pretrained(
        "black-forest-labs/FLUX.1-Fill-dev",
        transformer=transformer,
        torch_dtype=torch.bfloat16,
        trust_remote_code=True
    ).to(device)

    pipe.enable_model_cpu_offload()

Reproduction

https://huggingface.co/boricuapab/flux1-fill-dev-fp8/blob/main/README.md
https://huggingface.co/pengxian/diffusion_models/blob/main/flux1-fill-dev_fp8.safetensors

Logs

System Info

Windows 11
Python 11

Who can help?

No response

[DN6]

@SlimRG Could you share the error log you're seeing? Not sure what the issue is here? Is the model not loading?

[SlimRG]

t```
ransformer = FluxTransformer2DModel.from_single_file(
"https://huggingface.co/YarvixPA/FLUX.1-Fill-dev-gguf/blob/main/flux1-fill-dev-Q6_K.gguf",
quantization_config=GGUFQuantizationConfig(compute_dtype=torch.bfloat16),
torch_dtype=torch.bfloat16
).to("cuda")

pipe = FluxFillPipeline.from_pretrained(
    "black-forest-labs/FLUX.1-Fill-dev",
    transformer=transformer,
    torch_dtype=torch.bfloat16
).to("cuda")

#pipe.enable_model_cpu_offload()
return pipe