CuDNN failed to initialize

I installed TensorFlow 2.0 along with it came cudnn 7.6.4 installed with cuda toolkit 10.0.130. Now, whenever I run any script on TensorFlow following statement appears:

Unknown Error: Cudnn Failed to initialize see previous logs to check whether an error occurred.

So I checked the logs I got this:

[b]Invoking ptxas not supported on Windows
Relying on driver to perform ptx compilation. This message will be only logged once.
[/b]

What I understood from the link https://www.tensorflow.org/install/source#tested_build_configurations is that TensorFlow 2.0 runs on cudnn 7.4.

Does the compatibility issue is causing problems? If so how could I downgrade my cudnn on windows library?

I have used conda to install TensorFlow.

Hello,

I believe Tensorflow’s cuDNN dependency is >=7.4.1 https://www.tensorflow.org/install/gpu#software_requirements. So 7.6 should be fine.

If you’d like to downgrade cudnn, you’ll need to uninstall the cuda toolkit (http://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html#uninstall-cuda-software) and reinstall the version you’d prefer.

Thanks for the suggestion.

Can you please guide to resolve the following error occurring in Jupyter Notebook
Unknown Error: Cudnn Failed to initialize see previous logs to check whether an error occurred.

And regarding the downgrading suggestion, I didn’t installed cudnn 7.6.4 by myself it was installed automatically in the conda environment. The method you suggested won’t reflect in the conda environment I guess.

I got the same Error.
Anaconda
Tensorflow-gpu 2.0
CUDA 10
Any solution ? :)