本地加载生图模型，为啥这么慢，怎么多卡推理

Loading models from: [
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/transformer/diffusion_pytorch_model-00001-of-00005.safetensors",
"/appdata/models_download/Qwen/Qwen-Image-Edit-2511/transformer/diffusion_pytorch_model-00002-of-00005.safetensors",
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/transformer/diffusion_pytorch_model-00003-of-00005.safetensors",
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/transformer/diffusion_pytorch_model-00004-of-00005.safetensors",
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/transformer/diffusion_pytorch_model-00005-of-00005.safetensors"
]
Loaded model: {
"model_name": "qwen_image_dit",
"model_class": "diffsynth.models.qwen_image_dit.QwenImageDiT",
"extra_kwargs": null
}
Loading models from: [
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/text_encoder/model-00001-of-00004.safetensors",
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/text_encoder/model-00002-of-00004.safetensors",
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/text_encoder/model-00003-of-00004.safetensors",
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/text_encoder/model-00004-of-00004.safetensors"
]
Loaded model: {
"model_name": "qwen_image_text_encoder",
"model_class": "diffsynth.models.qwen_image_text_encoder.QwenImageTextEncoder",
"extra_kwargs": null
}
Loading models from: "/appdata//models_download/Qwen/Qwen-Image-Edit-2511/vae/diffusion_pytorch_model.safetensors"
Loaded model: {
"model_name": "qwen_image_vae",
"model_class": "diffsynth.models.qwen_image_vae.QwenImageVAE",
"extra_kwargs": null
}
Using qwen_image_text_encoder from [
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/text_encoder/model-00001-of-00004.safetensors",
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/text_encoder/model-00002-of-00004.safetensors",
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/text_encoder/model-00003-of-00004.safetensors",
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/text_encoder/model-00004-of-00004.safetensors"
].
Using qwen_image_dit from [
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/transformer/diffusion_pytorch_model-00001-of-00005.safetensors",
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/transformer/diffusion_pytorch_model-00002-of-00005.safetensors",
"/appdata/models_download/Qwen/Qwen-Image-Edit-2511/transformer/diffusion_pytorch_model-00003-of-00005.safetensors",
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/transformer/diffusion_pytorch_model-00004-of-00005.safetensors",
"/appdata//models_download/Qwen/Qwen-Image-Edit-2511/transformer/diffusion_pytorch_model-00005-of-00005.safetensors"
].
Using qwen_image_vae from "/appdata//models_download/Qwen/Qwen-Image-Edit-2511/vae/diffusion_pytorch_model.safetensors".
No qwen_image_blockwise_controlnet models available. This is not an error.
Downloading Model from https://www.modelscope.cn to directory: /appdata/zhangkailin/ML/models/Qwen/Qwen-Image
No siglip2_image_encoder models available. This is not an error.
No dinov3_image_encoder models available. This is not an error.
No qwen_image_image2lora_style models available. This is not an error.
No qwen_image_image2lora_coarse models available. This is not an error.
No qwen_image_image2lora_fine models available. This is not an error.

一直卡在这了，等了很久还没输出

Start by reproducing the Qwen-Image-Edit-2511 load using the logged model paths and inspect the transition after the ModelScope download message. Trace the qwen_image_dit, qwen_image_text_encoder, and qwen_image_vae loading entry points to determine whether the stall is downloading or initializing; done means identifying the blocking step and defining the supported multi-GPU inference path.