运行K采样器时出错

[JasonKEN432580]

compile model+用于flux depth redux 模型，K采样器如下报错，请问怎样解决

backend='inductor' raised:
FileExistsError: [WinError 183] 当文件已存在时，无法创建该文件。: 'C:\Users\JasonL\AppData\Local\Temp\torchinductor_JasonL\cache\.11940.26028.tmp' -> 'C:\Users\JasonL\AppData\Local\Temp\torchinductor_JasonL\cache\7f3baf1735ba64fd83f502c7daa8bfc367153e3749ad72b239e323c8d0abf62c'

Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information

You can suppress this exception and fall back to eager by setting:
import torch._dynamo
torch._dynamo.config.suppress_errors = True

[chengzeyi]

It seems that torch.compile has some problems running on your Windows. Could you try cleaning up your torchinductor cache file directory or try WSL2?

I personally use WSL2 to run and find it very stable and reliable.

[JasonKEN432580]

WSL2 means?

[chengzeyi]

WSL2 means?

windows subsystem for linux 2

[chengzeyi]

Seems duplicate of #59 and #32

[judian17]

也许是模型导致了这个错误？我可以使用flux模型编译并生成图片，但我使用尝试对sdxl模型编译时出现了与你一样的错误。也许你可以试试其他的编译节点，例如comfyui原生的torch compile或者kijai nodes里的编译节点。对于flux模型，kijai提供了一个叫做TorchCompileModelFluxAdvanced的节点，可以非常快速的完成编译。我不知道以上对于flux depth redux是否同样有效。Perhaps the model is causing this error? I can compile and generate images using the flux model, but I encounter the same error as you when I try to compile the sdxl model. Maybe you could try other compilation nodes, such as the native torch compile in comfyui or the compilation node in kijai nodes. For the flux model, kijai provides a node called TorchCompileModelFluxAdvanced, which can complete the compilation very quickly. I'm not sure if the above is equally effective for flux depth redux.

[uiueux]

Encounter the same error, whether on Windows or Linux.

[JasonKEN432580]

也许是模型导致了这个错误？我可以使用flux模型编译并生成图片，但我使用尝试对sdxl模型编译时出现了与你一样的错误。也许你可以试试其他的编译节点，例如comfyui原生的torch compile或者kijai nodes里的编译节点。对于flux模型，kijai提供了一个叫做TorchCompileModelFluxAdvanced的节点，可以非常快速的完成编译。我不知道以上对于flux depth redux是否同样有效。Perhaps the model is causing this error? I can compile and generate images using the flux model, but I encounter the same error as you when I try to compile the sdxl model. Maybe you could try other compilation nodes, such as the native torch compile in comfyui or the compilation node in kijai nodes. For the flux model, kijai provides a node called TorchCompileModelFluxAdvanced, which can complete the compilation very quickly. I'm not sure if the above is equally effective for flux depth redux.

I have tried TorchCompileModelFluxAdvanced,but it doesn't work~

[judian17]

@JasonKEN432580 你试着把报错里提到的两个临时文件清理后再试试？Try cleaning up the two temporary files mentioned in the error report and then give it another shot?

[skystoler]

it might be torch's issue, pytorch/pytorch#138211 and woct0rdho/triton-windows#10 might be helpful.