Closed Johk3 closed 1 year ago
You can create an anaconda environment and install tensorflow-gpu. It will install those it self.
conda create -n tensorflowgpu python=3.6 conda activate tensorflowgpu conda install tensorflow-gpu
then go to the requirements text and install everything except tensorflow-gpu==1.4.0 with pip by hand.
It seems like for some reason the tensorflow-gpu is looking for libcublas.so.9.0 which doesnt exist in cuda 9.1 that I have on my system. Would there be anyway around this?