I suspect a compatibility issue that is preventing tensorflow to use properly the GPU.
Nvidia Geforce MX150
Driver version 445.75
From the documentation (https://docs.nvidia.com/deeplearning/sdk/cudnn-support-matrix/index.html#cudnn-versions-763) for CUDA/cudnn versions I have, compatible driver is 418.39 (from 440 and on, higher CUDA and cudnn versions are needed)
As I prefer keeping the CUDA 10.1 version, I want to downgrade my GPU driver to r418.39, but I couldn’t find anyway to do that.
Thank you for your support