Flux训练的Lora无法工作，如何正确加载？

[DoiryCool]

import torch
from optimum.quanto import freeze, qfloat8, quantize
from diffusers import FluxPipeline, FluxTransformer2DModel
from PIL import Image
from safetensors.torch import load_file


transformer = FluxTransformer2DModel.from_single_file(
    "/models/Flux/FLUX.1-dev/flux1-dev-fp8.safetensors",
    torch_dtype=torch.float16,
)
pipe = FluxPipeline.from_pretrained(
    "/models/Flux/FLUX.1-dev",
    transformer=transformer,
    torch_dtype=torch.float16,
).to("cuda")


pipe.load_lora_weights("/models/output/bridge-arch-flux-000004.safetensors")

generator = torch.Generator().manual_seed(12345)

[liangcheng-1998]

我也是这样加载，和没加载生成的图一样