This tutorial will help you to run Tensorflow in Windows 11 to use GPU for modelling, with the latest Visual Studio, CUDA and latest drivers. In this case, I’m using RTX 3060Ti Lite Hashes and latest Python for Windows.
We need to install
- Clean Windows Installation and GPU drivers card
- the Microsoft Visual C++ (MSVC) compiler 2022
- the CUDA Toolkit 11.8
- the cuDNN libraries
- Install tensorflow
1. Windows fresh installation to remove the complexity of issues or bugs is recommended here. I did re-install my Windows 11 in this step
2. Instead of using Nvidia Geforce Experience, I uninstall the GPU drivers (back to Microsoft Basic) and use NVCleanstall to download my RTX 3060 drivers without others bloatware https://www.techpowerup.com/download/techpowerup-nvcleanstall/
3. I use Python 3.9 from Anaconda https://www.anaconda.com/products/distribution
We need to install Microsoft Visual Studio 2022, which it will required by CUDA for integration. You can download the community version and check for C/C++ development installation
Once its already installed in your desktop, then go for next step for install CUDA 11 (the latest version for Windows)
When you completed this step, its halfway done!
Download the CUDNN https://developer.nvidia.com/rdp/cudnn-download and use the latest version. Then extract the zip files and copy the contents following to
cuda\bin\cudnn*.dll to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\bin. cuda\include\cudnn*.h to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\include. cuda\lib\x64\cudnn*.lib to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\lib\x64.
Now, we are ready to install tensorflow
pip install tensorflow-gpu
Now, you can test the tensorflow now working using GPU RTX Nvidia.
python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"