Tensorflow on GPU supported on native-Windows

TensorFlow
,
1

Hi there,
I have a few questions. For installing tensorflow on Windows with NVIDIA GPU, it is mentioned:
conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0

It also says
TensorFlow 2.10 was the last TensorFlow release that supported GPU on native-Windows.
Starting with TensorFlow 2.11, you will need to install TensorFlow in WSL2

My question is whether TensorFlow 2.10 will work with an upgraded CUDA 12.4 and cuDNN SDK 8.9.

As for Tensorflow in WSL2, will the command: pip install tensorflow[and-cuda]
install the latest version of tensorflow, ie: 2.18? or it would depend on the CUDA/Python versions on WSL2?

Regards,
Amit Agarwal

3

Hi @agarwalamit081, It is recommended to use the test build configuration for the smooth installation.

By installing using this pip install tensorflow[and-cuda] command the latest version of tensorflow 2.18 and the cuda packages supported for tensorflow will be installed.

It would also depends on Python version. so you have to install the Python version as per the test build configuration. Thank You.

4

Thanks for the clarifications.