Procedure entry point fatBinaryCtl_CreateHandle not found in C:\WINDOWS\system32\nvcuda.dll

Gaming and Visualization Technologies General Topics and Other SDKs Drivers - Linux, Windows, MacOS
1

I have a GTX 1080 on Windows 10 with the latest drivers and CUDA 10. After installing pytorch via Anaconda, I try to execute torch.cuda.is_available() in the python console but I get the above error. Anybody has run into a similar problem or knows what to do? I’ve reinstalled the graphics drivers, CUDA and the Python packages countless times, to no avail.

numba -s does list that it has detected the GPU and CUDA so not sure what’s wrong:

__CUDA Information__ Found 1 CUDA devices id 0 b'GeForce GTX 1080' [SUPPORTED] compute capability: 6.1 pci device id: 0 pci bus id: 1 Summary: 1/1 devices are supported CUDA driver version : 10010
2

HI,

I’m sorry it is not an answer… I just wanted to mention that I am getting exactly the same error. Just that in my case I used Miniconda, and created a new environment for PyTorch. and installed the latest PyTorch in it. [conda install pytorch torchvision cudatoolkit=10.0 -c pytorch]. My env is the following:

>conda list # packages in environment at C:\Users\gmseabra\Miniconda3\envs\pytorch_1.0: # # Name Version Build Channel asn1crypto 0.24.0 py37_0 blas 1.0 mkl ca-certificates 2019.1.23 0 certifi 2018.11.29 py37_0 cffi 1.11.5 py37h74b6da3_1 chardet 3.0.4 py37_1 console_shortcut 0.1.1 3 cryptography 2.5 py37h7a1dbc1_0 cudatoolkit 10.0.130 0 freetype 2.9.1 ha9979f8_1 icc_rt 2019.0.0 h0cc432a_1 idna 2.8 py37_0 intel-openmp 2019.1 144 jpeg 9b hb83a4c4_2 libpng 1.6.36 h2a8f88b_0 libtiff 4.0.10 hb898794_2 menuinst 1.4.14 py37hfa6e2cd_0 mkl 2019.1 144 mkl_fft 1.0.10 py37h14836fe_0 mkl_random 1.0.2 py37h343c172_0 ninja 1.8.2 py37he980bc4_1 numpy 1.15.4 py37h19fb1c0_0 numpy-base 1.15.4 py37hc3f5095_0 olefile 0.46 py37_0 openssl 1.1.1a he774522_0 pillow 5.4.1 py37hdc69c19_0 pip 19.0.1 py37_0 pycosat 0.6.3 py37hfa6e2cd_0 pycparser 2.19 py37_0 pyopenssl 19.0.0 py37_0 pysocks 1.6.8 py37_0 python 3.7.2 h8c8aaf0_2 pytorch 1.0.1 py3.7_cuda100_cudnn7_1 pytorch pywin32 223 py37hfa6e2cd_1 requests 2.21.0 py37_0 ruamel_yaml 0.15.46 py37hfa6e2cd_0 setuptools 40.8.0 py37_0 six 1.12.0 py37_0 sqlite 3.26.0 he774522_0 tk 8.6.8 hfa6e2cd_0 torchvision 0.2.1 py_2 pytorch urllib3 1.24.1 py37_0 vc 14.1 h0510ff6_4 vs2015_runtime 14.15.26706 h3a45250_0 wheel 0.32.3 py37_0 win_inet_pton 1.0.1 py37_1 wincertstore 0.2 py37_0 xz 5.2.4 h2fa13f4_4 yaml 0.1.7 hc54c509_2 zlib 1.2.11 h62dcd97_3 zstd 1.3.7 h508b16e_0

I hope anyone else there has an answer…

3

Hi,

I have the exact same issue on Windows with cuda 10 and the latest drivers. I believe this is an issue with pytorch and not CUDA. The other GPU enabled packages in my environment (catboost, numba) work fine. I fixed the issue by uninstalling NVIDIA drivers and miniconda, reverting to a previous system state (using system restore), and reinstalling everything again.

4

When you did that, did you reinstall exactly the same versions you had before? Or did you have to rollback any library to an earlier version?

5

I installed it on the current version of miniconda3, which is available on their website. All my package version correspond to whatever’s available on the current version (without doing any updates). If I updated using conda update --all, the problem comes back. The only solution for now (until pytorch fixes this) is to install pytorch in a new environment.

6

Thanks! I confirm that this works:

  1. Uninstall Miniconda
  2. Reinstall Miniconda, but do not update
  3. Create a new env for PyTorch
  4. Install PyTorch in this new env

Now the

>>> torch.cuda.is_available() True

Works. However, in interactive shells, I’m back to the issue I described at https://github.com/pytorch/pytorch/issues/17108#issuecomment-463901596, where the

torch.cuda.current_device()

command works when executed from a file but does not work from an interactive shell. I’ll open a new topic for this one.

Thanks!

7

Thanks for your replies folks. Yes, it seems to be a pytorch issue and I’ll post my problem in their support forums/bug tracker.

8

Notice that this issue has already been raised there this afternoon:

[url]https://github.com/pytorch/pytorch/issues/17233[/url]

9

Thanks. I’ll add a +1 to that issue.