Old way of Tensorflow installation @ linux


I have to install TF2 + a few more things at work (which limits my options)
Unfortunately, the new method does not work correctly for me (i.e. ``pip install tensorflow[and-cuda]‘’ )

How I can find on the web page the description of the “old way” of tf2 installation,
i.e. with miniconda, separate installation of cuda, cudnn then setting the path and finally to the two and finally installing TF2 (+ eventually fixing ubuntu problem…)

Thanks in advance for your help!

Hi @mstolars

Welcome to the TensorFlow Forum!

The new way is more efficient to install the TensorFlow for GPU support along with it’s correct supported version of CUDA and cuDNN libraries. There is no need to check and to install separately the compatible version of these libraries to it’s respective TensorFlow version and to set the path for each.

Could you please try again the same command lines by checking the Hardware/Software requirements to install the TensorFlow with GPU setup in your system and let us know if the issue still persists. Thank you.