ImportError: DLL load failed while importing bias_act_plugin

[ziyuwzf]

Describe the bug
ImportError: DLL load failed while importing bias_act_plugin: 找不到指定的模块。

To Reproduce
Steps to reproduce the behavior:

1. In 'stylegan3-main' directory, run command 'python gen_images.py', and before it i set "F:\Python39\lib\site-packages\torch\utils\cpp_extension.py" line 1682 code from "command = ['ninja', '-v']" to " command = ['ninja', '--version']"
2. See error (copy&paste full log, including exceptions and stacktraces).
   F:\Python39\python.exe F:/PycharmProjects/stylegan3-main/gen_images.py --outdir=out --trunc=1 --seeds=2 --network=./files/stylegan3-r-afhqv2-512x512.pkl
   Loading networks from "./files/stylegan3-r-afhqv2-512x512.pkl"...
   Generating image for seed 2 (0/1) ...
   Setting up PyTorch plugin "bias_act_plugin"... Failed!
   Traceback (most recent call last):
   File "F:\PycharmProjects\stylegan3-main\gen_images.py", line 143, in
   generate_images() # pylint: disable=no-value-for-parameter
   File "F:\Python39\lib\site-packages\click\core.py", line 1128, in call
   return self.main(*args, **kwargs)
   File "F:\Python39\lib\site-packages\click\core.py", line 1053, in main
   rv = self.invoke(ctx)
   File "F:\Python39\lib\site-packages\click\core.py", line 1395, in invoke
   return ctx.invoke(self.callback, **ctx.params)
   File "F:\Python39\lib\site-packages\click\core.py", line 754, in invoke
   return __callback(*args, **kwargs)
   File "F:\PycharmProjects\stylegan3-main\gen_images.py", line 135, in generate_images
   img = G(z, label, truncation_psi=truncation_psi, noise_mode=noise_mode)
   File "F:\Python39\lib\site-packages\torch\nn\modules\module.py", line 1102, in _call_impl
   return forward_call(*input, **kwargs)
   File "", line 503, in forward
   File "F:\Python39\lib\site-packages\torch\nn\modules\module.py", line 1102, in _call_impl
   return forward_call(*input, **kwargs)
   File "", line 143, in forward
   File "F:\Python39\lib\site-packages\torch\nn\modules\module.py", line 1102, in _call_impl
   return forward_call(*input, **kwargs)
   File "", line 92, in forward
   File "F:\PycharmProjects\stylegan3-main\torch_utils\ops\bias_act.py", line 84, in bias_act
   if impl == 'cuda' and x.device.type == 'cuda' and _init():
   File "F:\PycharmProjects\stylegan3-main\torch_utils\ops\bias_act.py", line 41, in _init
   _plugin = custom_ops.get_plugin(
   File "F:\PycharmProjects\stylegan3-main\torch_utils\custom_ops.py", line 140, in get_plugin
   torch.utils.cpp_extension.load(name=module_name, build_directory=cached_build_dir,
   File "F:\Python39\lib\site-packages\torch\utils\cpp_extension.py", line 1124, in load
   return _jit_compile(
   File "F:\Python39\lib\site-packages\torch\utils\cpp_extension.py", line 1362, in _jit_compile
   return _import_module_from_library(name, build_directory, is_python_module)
   File "F:\Python39\lib\site-packages\torch\utils\cpp_extension.py", line 1752, in _import_module_from_library
   module = importlib.util.module_from_spec(spec)
   File "(frozen importlib._bootstrap)", line 565, in module_from_spec
   File "(frozen importlib._bootstrap_external)", line 1173, in create_module
   File "(frozen importlib._bootstrap)", line 228, in _call_with_frames_removed
   ImportError: DLL load failed while importing bias_act_plugin: 找不到指定的模块。

Expected behavior
A clear and concise description of what you expected to happen.

Desktop (please complete the following information):

• OS: [Windows 10]
• PyTorch version (pytorch 1.10.0+cu113)
• CUDA toolkit version (CUDA 11.4)
• NVIDIA driver version（471）
• GPU [RTX 1060]

i have tried remove C:\Users{username}\AppData\Local\torch_extensions\torch_extensions\Cache\bias_act_plugin,but it not works

[WeiDuan123]

Have you solved the problem? I have met the same question.

[bmquynhlinh]

I met this problem and was able to solve it:

1. Check whether you install ninja or not
2. I read the error logs and found out that ninja tries to load my CUDA 10.0 instead of the 11.2 version. So, I uninstall CUDA 10.0 and then bias_act_plugin loaded successfully

[gengzheliu]

i have exactly the same problems with you. i treid uninstall CUDA11.3 and pytorch1.10.2, but still doesn't work. is that problem with the verison?

[gengzheliu]

• Check whether you install ninja or not
• I read the error logs and found out that ninja tries to load my CUDA 10.0 instead of the 11.2 version. So, I uninstall CUDA 10.0 and then bias_act_plugin loaded successfully

Hi is that mean you installed 2 verison of CUDA and then you uninstalled one of them?

[bmquynhlinh]

Yes, I uninstalled it because I don't need it anymore. But, if you can keep multiple CUDA versions, and assign 11.2 to CUDA_HOME in env. It still works

[ziyuwzf]

I met this problem and was able to solve it:

1. Check whether you install ninja or not
2. I read the error logs and found out that ninja tries to load my CUDA 10.0 instead of the 11.2 version. So, I uninstall CUDA 10.0 and then bias_act_plugin loaded successfully

before CUDA11.4，i installed CUDA10,x，but i have assigned CUDA11.4 to CUDA_HOME in env，and no 10.x in env path.i will try you advisement,thanks

[49xxy]

I have the same problem. Have you solved it？Looking forward to your reply！

[jannehellsten]

I recently found that VS2019 community edition doesn't work with StyleGAN3 but VS2022 does. Alas, using VS2022 required some small changes. In case anyone above is struggling, this change might help:

69c7ef0

[joinforcookies]

I had CUDA 12.1. I uninstalled it and installed CUDA 11.8 from https://developer.nvidia.com/cuda-11-8-0-download-archive?target_os=Windows&target_arch=x86_64&target_version=11&target_type=exe_local. How did I know? It was based on the available CUDA versions available in this PyTorch website https://pytorch.org/get-started/locally/ and it solved the issue.

[toprakfirat]

I had CUDA 12.1. I uninstalled it and installed CUDA 11.8 from https://developer.nvidia.com/cuda-11-8-0-download-archive?target_os=Windows&target_arch=x86_64&target_version=11&target_type=exe_local. How did I know? It was based on the available CUDA versions available in this PyTorch website https://pytorch.org/get-started/locally/ and it solved the issue.

I had the same issue, this worked for me.

[syf2012]

I had CUDA 12.1. I uninstalled it and installed CUDA 11.8 from https://developer.nvidia.com/cuda-11-8-0-download-archive?target_os=Windows&target_arch=x86_64&target_version=11&target_type=exe_local. How did I know? It was based on the available CUDA versions available in this PyTorch website https://pytorch.org/get-started/locally/ and it solved the issue.

May I ask what version of torch you are using? I also installed cuda11.8, but it still does not work

[Tempesta-96]

Hello, does the version of ninja matter? The logs mentioned ninja_required_version = 1.3 but I'm using ninja==1.11.1.3 cant seem to downgrade my ninja version too.

I'm using ninja==1.11.1.3
torch==1.9.0+cu111
torchaudio==0.9.0
torchvision==0.10.0+cu111

and my ninja is pointing to CUDA 11.1 nvcc = C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\bin\nvcc