DiffSynth Studio 无GPU模型下载和运行调式-CSDN博客

本文链接：https://blog.csdn.net/cybersnow/article/details/147024954

  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\.glut\lib\site-packages\streamlit\runtime\scriptrunner\script_runner.py", line 542, in _run_script
    exec(code, module.__dict__)
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\pages\2_Video_Creator.py", line 197, in <module>
    SDVideoPipelineRunner(in_streamlit=True).run(config)
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\diffsynth\pipelines\stable_diffusion_video.py", line 340, in run
    model_manager, pipe = self.load_pipeline(**config["models"])
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\diffsynth\pipelines\stable_diffusion_video.py", line 273, in load_pipeline
    pipe = SDVideoPipeline.from_model_manager(
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\diffsynth\pipelines\stable_diffusion_video.py", line 116, in from_model_manager
    pipe.fetch_main_models(model_manager)
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\diffsynth\pipelines\stable_diffusion_video.py", line 82, in fetch_main_models
    self.text_encoder = model_manager.text_encoder
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\diffsynth\models\__init__.py", line 237, in __getattr__
    return super.__getattribute__(__name)
TypeError: descriptor '__getattribute__' requires a 'super' object but received a 'str'
2025-04-06 03:39:09.795 Uncaught app exception

解释：需要一个模型，但是你没有

ypeError: descriptor '__getattribute__' requires a 'super' object but received a 'str'

开始下载模型

继续运行

aceback (most recent call last):
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\.glut\lib\site-packages\streamlit\runtime\scriptrunner\script_runner.py", line 542, in _run_script
    exec(code, module.__dict__)
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\pages\2_Video_Creator.py", line 197, in <module>
    SDVideoPipelineRunner(in_streamlit=True).run(config)
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\diffsynth\pipelines\stable_diffusion_video.py", line 340, in run
    model_manager, pipe = self.load_pipeline(**config["models"])
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\diffsynth\pipelines\stable_diffusion_video.py", line 272, in load_pipeline
    model_manager.load_models(model_list, lora_alphas=lora_alphas)
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\diffsynth\models\__init__.py", line 211, in load_models
    self.load_model(file_path, lora_alphas=lora_alphas)
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\diffsynth\models\__init__.py", line 201, in load_model
    self.load_sd_lora(state_dict, alpha=lora_alphas.pop(0))
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\diffsynth\models\__init__.py", line 151, in load_sd_lora
    SDLoRA().add_lora_to_text_encoder(self.model["text_encoder"], state_dict, alpha=alpha, device=self.device)
KeyError: 'text_encoder'

解释历史lora错误

model_manager.load_models(model_list, lora_alphas=lora_alphas)
File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\diffsynth\models\__init__.py", line 211, in load_models
self.load_model(file_path, lora_alphas=lora_alphas)

修改源码显示加载模型，方便定位出问题模型

所在路径diffsynth\models\__init__.py 191行

 def load_model(self, file_path, components=None, lora_alphas=[]):
        print(f"未来之窗 cyberwin model_name: {file_path} now ")
        state_dict = load_state_dict(file_path, torch_dtype=self.torch_dtype)
        if self.is_animatediff(state_dict):

再次运行视频制作

cyberwin model_name: models/stable_diffusion\stableDiffusion35VAE_official.safetensors now
cyberwin model_name: models/AnimateDiff\animatediffV3Models_adapterFP32.safetensors now
cyberwin model_name: models/lora\loraPeaceSign_v03.safetensors now
2025-04-06 14:11:51.436 Uncaught app exception
Traceback (most recent call last):

解释模型 stableDiffusion35VAE_official 正常 animatediffV3Models_adapterFP32

loraPeaceSign_v03 但是lora错误了

立即更换lora模型

如何更换lora模型

搜索模型，并下载

lora模型运行监测-通过

cyberwin model_name: models/stable_diffusion\stableDiffusion35VAE_official.safetensors now
cyberwin model_name: models/AnimateDiff\animatediffV3Models_adapterFP32.safetensors now
cyberwin model_name: models/lora\lora_rioko.safetensors now
cyberwin model_name: models/ControlNet\15.safetensors now
cyberwin model_name: models/ControlNet\15.safetensors now
cyberwin model_name: models/ControlNet\15.safetensors now
2025-04-06 14:17:37.875 Uncaught app exception
Traceback (most recent call last):
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\.glut\lib\site-packages\streamlit\runtime\scriptrunner\script_runner.py", line 542, in _run_script
    exec(code, module.__dict__)
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\pages\2_Video_Creator.py", line 197, in <module>
    SDVideoPipelineRunner(in_streamlit=True).run(config)
  File "E:\ai_proe\DiffSynthStudio3g\DiffSynth-Studio\diffsynth\pipelines\stable_diffusion_video.py", line 340, in run
    model_manager, pipe = self.load_pipeline(*

controlnet 模型报错

controlnet 模型报错版本解释

softedge，软边

openpose 开放姿态

lineart 线性

lineart_anime 线性动漫

canny 精明

depth 深度

tile 瓦

术语	含义	在 ControlNet 中的作用	应用场景举例
Softedge（软边）	提取图像边缘信息，边缘更柔和、连续，含更多细节和过渡	帮助模型捕捉物体形状和轮廓，使生成图像边缘自然流畅	风景图像生成，如勾勒山脉、河流边界
OpenPose（开放姿态）	人体姿态估计模型，检测人体关键点获取姿态信息	控制生成图像中人物姿态，确保肢体动作自然协调	动画制作、游戏角色设计中特定动作人物图像生成
Lineart（线性）	提取图像线条信息，简化为线条表示，突出物体轮廓和结构	为模型提供图像基本架构指导，用于生成以线条结构为主的插画	平面设计、漫画创作
Lineart_anime（线性动漫）	针对动漫风格图像生成的线性控制模式，线条更简洁、夸张，突出动漫风格特征	引导模型生成具有典型动漫风格线条的图像	动漫角色、场景图像创作
Canny（Canny 边缘检测）	经典边缘检测算法，明确图像中物体边缘位置	模型依据其检测出的边缘信息生成完整图像，保证物体边缘准确	建筑设计、机械制图等对边缘准确性要求高的图像生成
Depth（深度）	图像中物体与相机（或观察者）之间的距离关系，即深度信息	为模型提供图像三维空间结构线索，生成具有真实远近层次和空间感的图像	虚拟现实、3D 场景生成
Tile（瓦片）	将输入图像视为可重复的 “瓦片” 结构	生成具有连续、重复图案且拼接处自然过渡的图像	壁纸、织物纹理等需要重复图案的图像生成，如室内设计、纺织品设计