Could not find 'cudart64_100.dll'

Ok, this happens when I install TensorFlow-GPU via the file> settings> project interpreter> + <add new library>.  Now it redirects me to a link:

ImportError: Could not find 'cudart64_100.dll'. TensorFlow requires that this DLL be installed in a directory that is named in your %PATH% environment variable. Download and install CUDA 10.0 from this URL: https://developer.nvidia.com/cuda-90-download-archive 

This error message is great. However, the cuda-90 archive does not, have a file 'cudart64_100.dll', although I did only check about 10-15 of the archives. So I went to the 10.1 version and it has curdart64_101.dll which is not helpful.  Does anyone know a workaround as I am not a huge fan of colab.

6 comments
Comment actions Permalink

Hi,

Do you also get this error when running your script from the terminal outside of IDE? Just telling, because it seems like something unrelated to PyCharm and you could also ask on TensorFlow forums...

0
Comment actions Permalink

I see what you mean about it being most likely un-related to PyCharm.

> python app.py

ImportError: Could not find 'cudart64_90.dll'. TensorFlow requires that this DLL be installed in a directory that is named in your %PATH% environment variable. Download and install CUDA 9.0 from this URL:http:// https://developer.nvidia.com/cuda-toolkit . Yes, I did uninstall CUDA 9.0 was hoping to get away with not having it, yet I can tell you for sure that it does not have the cudart64_90.dll file.

I have a funny feeling that Nvida has stopped access to the the CUDA 10.0 and will try on TensorFlow or Nvida forums and will get back to you.  Incidentally, Stake OverFlow also points there users to the http:// https://developer.nvidia.com/cuda-toolkit.

0
Comment actions Permalink

Well how embarrassing, I don't know where I ended up on the NVidia website, yet it was not at the address below.  Nevertheless, I think I have found it.

https://developer.nvidia.com/cuda-toolkit-archive 

0
Comment actions Permalink

For anyone else having trouble with installing cuda, you will need to also install the CUDA Toolkit 9.0 and also download the Download cuDNN v7.4.1 (Nov 8, 2018), for CUDA 9.0 which, has the cudnn64_7.dll file in the /bin folder (see image below).  The cudnn64_7.dll then needs to be manually copied to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0\bin. For further information and help such as setting up environment paths, see this blog https://medium.com/@teavanist/install-tensorflow-gpu-on-windows-10-5a23c46bdbc7 

0
Comment actions Permalink

Hey guys,

Same issues I have been gone through here  "Could not find 'cudart64_100.dll'. TensorFlow requires that this DLL be installed in a directory that is named in your %PATH% environment variable. Download and install CUDA 10.0 from this URL: https://developer.nvidia.com/cuda-90-download-archive"

this issues can be solved by reinstallation only  and never update the Tensorflow because it leads to further problem in numpy " giving error ModuleNotFoundError: No module named 'numpy.core._multiarray_umath' ImportError: numpy.core.multiarray failed to import  whenever we update i know that extra features get added up but those feature are not supported by another set(Cuda ) they are interlinked and depended on each other ,as in case on Cuda 9.0 and Cuda 10.0 they have 'cudart64_90.dll  and 'cudart64_100.dll are required . "which leads change in cuda due to which tensorflow gives a call "Could not find 'cudart64_100.dll".They only way is reinstallation .

This are the error I got in this flow which is given above ,So this are the issues I have faced. After reinstalling it is solved .

I think this could help you ,

Thank you  

0
Comment actions Permalink

To fix the issue, either:

  • Install CUDA from September 2018 which comes with 'cudart64_100.dll' library

Or, if you want the future TensorFlow support:

  • Install the latest ‘NVIDIA GPU Computing Toolkit’.
  • However, because the newest version of the CUDA library is 'cudart64_101.dll' and TensorFlow 2.0 requires the older 'cudart64_100.dll', also install the CUDA from September 2018.
  • Then copy 'cudart64_100.dll' library from the old install of CUDA to the new one.
  • Or just download the library from the file attached further below and copy it into your last CUDA installation directory in program files.

 

Step by step instructions on how to resolve the problem:

https://www.joe0.com/2019/10/19/how-resolve-tensorflow-2-0-error-could-not-load-dynamic-library-cudart64_100-dll-dlerror-cudart64_100-dll-not-found/

0

Please sign in to leave a comment.