No cudart64_110.dll and C:\Program Files\NVIDIA GPU Computing Toolkit\ directory

General Discussion
,
#1

I’m getting an error “Could not load dynamic library ‘cudart64_110.dll’; dlerror: cudart64_110.dll not found”
I have installed TensorFlow step by step with the official TensorFlow install tutorial.
I’m 100% sure I made every single instruction correctly. I have installed Microsoft Visual C++ Redistributable from link in the tutorial. I have installed Miniconda from direct link in that tutorial. I have the latest NVIDIA driver (522.25) and GTX 1050 ti GPU.
I’m sure I have done 5th and every next step with my virtual environment activated (conda activate tf). The 7th steps ran correctly but with “python” instead of “python3”. When I try with “3” it says there is no such file. I’m using Windows 10 21H2.
For some reason, I even don’t have C:\Program Files\NVIDIA GPU Computing Toolkit\ directory.
I have tried to repeat the 5th step (conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0) and it says “All requested packages already installed.”.
Path I have added to my “Path”:
C:\ProgramData\Miniconda3\Library\bin
C:\ProgramData\Miniconda3
C:\ProgramData\Miniconda3\Scripts

Sorry for my English. It’s not my primary language.

#3

Can you tell us what version of Tensorflow you are trying to install in Windows?

Quick fix is to download ‘cudart64_110.dll’ from here and paste it into cudatoolkit/bin file location. Thank you.

#4

I’m using Tensorflow 2.10. Where is cudatoolkit/bin located? I noticed that this problem only occurs in Visual Studio Code and it works fine in Anaconda Prompt. I have tried to copy cudart64_110.dll to C:\ProgramData\Miniconda3\Library\bin and now there is no error about cudart64_110.dll but there are now more issues with other files after running some TensorFlow function for example test = tf.constant([1,2,3])

2022-10-14 17:15:51.844313: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cublas64_11.dll'; dlerror: cublas64_11.dll not found
2022-10-14 17:15:51.848554: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cublasLt64_11.dll'; dlerror: cublasLt64_11.dll not found
2022-10-14 17:15:51.853492: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cufft64_10.dll'; dlerror: cufft64_10.dll not found
2022-10-14 17:15:51.858641: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'curand64_10.dll'; dlerror: curand64_10.dll not found
2022-10-14 17:15:51.862961: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cusolver64_11.dll'; dlerror: cusolver64_11.dll not found
2022-10-14 17:15:51.867683: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cusparse64_11.dll'; dlerror: cusparse64_11.dll not found
2022-10-14 17:15:51.872853: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cudnn64_8.dll'; dlerror: cudnn64_8.dll not found
2022-10-14 17:15:51.876600: W tensorflow/core/common_runtime/gpu/gpu_device.cc:1934] Cannot dlopen some GPU libraries. Please make sure the missing libraries mentioned above are installed properly if you would like to use GPU. Follow the guide at https://www.tensorflow.org/install/gpu for how to download and setup the required libraries for your platform.
Skipping registering GPU devices...
#5

You can see that your Tensorflow installation is looking for CUDA libraries of version 11 and 10. So in order to fix this, install the valid combinations of CUDA, Tensorflow and cuDNN as per here and also refer to Installing cuDNN on Windows. Thank you

#7

May be cuda and cudnn libraries not installed properly on base that why its not showing you the directly. Make sure if it is installed proper on base or not.

#9

I have already installed TensorFlow, CUDA and cuDNN, but for some reason its files are only detected when testing TensorFlow directly from Anaconda Console. For some reason it does not work when executed, for example from Visual Studio Code. Should I add something to system path? I made a virtual environment named “tf” as official guide said. Should I install TensorFlow, CUDA and cuDNN in base environment?