DiffSynth-Studio低显存示例代码中的vram_config参数与ModelConfig不兼容导致报错

非常感谢各位大佬做出的贡献，我想尝试在4070Super的显卡推理Qwen-Image-Edit-2509模型，为了节省显存，不至于显存爆炸而使用https://github.com/modelscope/DiffSynth-Studio/blob/main/examples/qwen_image/model_inference_low_vram/Qwen-Image-Edit-2509.py提供的样例代码，直接复制粘贴到pycharm后进行推理但是报错了。我查看了ModelConfig的源码，发现vram_config配置的参数，有很多参数均是无效的，vram_config参数如下：

vram_config = {
    "offload_dtype": "disk",
    "offload_device": "disk",
    "onload_dtype": torch.float8_e4m3fn,
    "onload_device": "cpu",
    "preparing_dtype": torch.float8_e4m3fn,
    "preparing_device": "cuda",
    "computation_dtype": torch.bfloat16,
    "computation_device": "cuda",
}

ModelConfig如下：

  path: Union[str, list[str]] = None
    model_id: str = None
    origin_file_pattern: Union[str, list[str]] = None
    download_resource: str = "ModelScope"
    offload_device: Optional[Union[str, torch.device]] = None
    offload_dtype: Optional[torch.dtype] = None
    local_model_path: str = None
    skip_download: bool = False

具体报错如下：

File "E:\Project_Style\Qwen-Image-Edit\Project\demo.py", line 25, in <module>
    ModelConfig(model_id="Qwen/Qwen-Image-Edit-2509", origin_file_pattern="transformer/diffusion_pytorch_model*.safetensors", **vram_config),
TypeError: ModelConfig.__init__() got an unexpected keyword argument 'onload_dtype'

我的环境中diffsynth 的Version: 1.1.9，我不知道哪个版本才可以运行官方提供的代码，如果有知道的朋友，劳烦您告诉我一声，非常的感谢！
后来为了解决该问题，并且不明白为什么使用Qwen-Image-Edit-2509进行推理，却还要加载Qwen-Image的权重文件，我在modelscope中提前下载好Qwen-Image-Edit-2509模型。尝试修改代码，但依旧无效并报错。
具体代码如下：

vram_config = {
    "offload_dtype": "disk",
    "offload_device": "disk",}

model_path=r"E:\Project_Style\Qwen-Image-Edit\Models"
model_id="Qwen/Qwen-Image-Edit-2509"
pipe = QwenImagePipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id=model_id,
                    origin_file_pattern="transformer/",
                    local_model_path=model_path,
                    skip_download=True,
                    **vram_config
                    ),
        ModelConfig(model_id=model_id,
                    origin_file_pattern="text_encoder/",
                    local_model_path=model_path,
                    skip_download=True,
                    **vram_config
                    ),
        ModelConfig(model_id=model_id,
                    origin_file_pattern="vae/",
                    local_model_path=model_path,
                    skip_download=True,
                    **vram_config
                    ),],
    processor_config=ModelConfig(model_id=model_id,
                                 origin_file_pattern="processor/",
                                 skip_download=True,
                                 local_model_path=model_path,),

)

报错如下：

Traceback (most recent call last):
  File "E:\Project_Style\Qwen-Image-Edit\Project\降低显存的方式.py", line 28, in <module>
    pipe = QwenImagePipeline.from_pretrained(
  File "D:\Miniconda3\envs\llm\lib\site-packages\diffsynth\pipelines\qwen_image.py", line 331, in from_pretrained
    model_manager.load_model(
  File "D:\Miniconda3\envs\llm\lib\site-packages\diffsynth\models\model_manager.py", line 408, in load_model
    if model_detector.match(file_path, state_dict):
  File "D:\Miniconda3\envs\llm\lib\site-packages\diffsynth\models\model_manager.py", line 165, in match
    if len(state_dict) == 0:
TypeError: object of type 'NoneType' has no len()
Loading models from: E:\Project_Style\Qwen-Image-Edit\Models\Qwen/Qwen-Image-Edit-2509\transformer/

为了解决这些问题，我耗费了很长的时间依旧没有头绪，有朋友知道如何解决该问题的，我非常的愿意向您请教，在此提前向您表示感谢！

Start with examples/qwen_image/model_inference_low_vram/Qwen-Image-Edit-2509.py and compare its ModelConfig fields with the installed DiffSynth version 1.1.9. Then trace the loading path through qwen_image.py and model_manager.py, using the reported errors and local model path as reproduction points. Done means the example or documented local-model configuration runs without the reported TypeErrors.