How to install CUDA 11.2 and cuDNN 8.1 on WSL2 for TensorFLow 2?

Question

I am very much struggling to get NVIDIA's CUDA 11.2 and cuDNN 8.1 for TensorFlow 2 on my WSL2 machine.

• I have followed their guide to install the latest CUDA 11.8 version and it works. The guide can be found here and is the standard guide recommended by NVIDIA.
• Though, TF2 only supports certain CUDA/cuDNN combinations as shown here
• Further, I have found the index of previously available WSL2 releases of CUDA and 11.2.0-1 is listed! You can see it for yourself here

Can anybody give me the commands to get CUDA 11.2 and cuDNN 8.1 on my WSL2 machine so that I can finally get back to coding in TensorFlow 2? I am thinking about something like this :

wget https://developer.download.nvidia.com/compute/cuda/repos/wsl-ubuntu/x86_64/cuda-wsl-ubuntu.pin
sudo mv cuda-wsl-ubuntu.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/11.8.0/local_installers/cuda-repo-wsl-ubuntu-11-8-local_11.8.0-1_amd64.deb
sudo dpkg -i cuda-repo-wsl-ubuntu-11-8-local_11.8.0-1_amd64.deb
sudo cp /var/cuda-repo-wsl-ubuntu-11-8-local/cuda-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update
sudo apt-get -y install cuda

My specs :

• OS : Windows 10 (WSL2 compatible 22H2 release, the latest one)
• Processor : AMD Ryzen 7 5800H
• RAM : 32GB
• GPU : NVIDIA GeForce RTX 3070 Laptop GPU
• WSL2 Distro : Ubuntu 20.04.5 LTS

Answer 1

You could check out Nvidia's download installer for WSL-Ubuntu from its download archives.

wget https://developer.download.nvidia.com/compute/cuda/repos/wsl-ubuntu/x86_64/cuda-wsl-ubuntu.pin
sudo mv cuda-wsl-ubuntu.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/11.2.0/local_installers/cuda-repo-wsl-ubuntu-11-2-local_11.2.0-1_amd64.deb
sudo dpkg -i cuda-repo-wsl-ubuntu-11-2-local_11.2.0-1_amd64.deb
sudo apt-key add /var/cuda-repo-wsl-ubuntu-11-2-local/7fa2af80.pub
sudo apt-get update
sudo apt-get -y install cuda

Answer 2

Here is how I did it:

Follow the instructions from NVIDIA's CUDA Toolkit Downloads page to setup their package repository with the versions specifically for WSL2

> wget https://developer.download.nvidia.com/compute/cuda/repos/wsl-ubuntu/x86_64/cuda-keyring_1.1-1_all.deb
> sudo dpkg -i cuda-keyring_1.1-1_all.deb
> sudo apt-get update

And now the fun part, since the repo contains all previous CUDA versions too:

> sudo apt install cuda-toolkit-11-7

I hope this helps some fellow traveler ^^

Answer 3

I just read your question please follow.

1. Do you installed compatible Cuda versions with your target GPU device? Example: I use CuDa 11.6 and for the NVIDIA GeForce RTX 3070 Laptop GPU, 5120 compatible since 8.6, read from CuDA Nvidia GPU developer
2. Install Tensorflow then Tensorflow GPU, if it still does not recover your GPU you need to build it on your machine that is not a complex task but finding one existing on websites is a more convenient way Test build configuration
3. If you try to build it on Linux you need to install a compatibility version of Cuda and CuDNN Compatibility Matrix
4. Last I read this question multiple of times, you still can download target versions of CuDA archived CuDA and CuDNN archive link

That is because they question then screenshot added see TF2.11 and WSL2.0 related to my answers