tensorflow，tensorflow-gpu 无法识别 GPU

Question

this might be a repeated post but none of the other posts worked for me. i have tensorflow 1.14.0 and tensorflow-gpu 1.14.0 i have already installed the compatible Cuda and Cudnn versions required, and specified their path in environment variables. but still the tensorflow and/or tensorflow-gpu, don't recognize my GPU. so the of the validating code:

from tensorflow.python.client import device_lib
def get_available_devices():
    local_device_protos = device_lib.list_local_devices()
    return [x.name for x in local_device_protos]

print(get_available_devices()) 

is:

['/device:CPU:0']

OP: windows 10

IDE: Jupyter (anaconda)

GPU version: GTX 960M

note: some posts mention that you have to uninstall tensorflow at the end, but when i do, the code above doesn't even work...

Answer 1

You should definitely uninstall tensorflow from your system, the code not working after that step is a sign your tensorflow-gpu installation has some problems. This accepted answer suggests uninstalling tensorflow via pycharms and installing via pip.

Also if you have the opportunity, try to upgrade to tensorflow >= 2.0, the eager execution is very nice and surely pays up the hassle of getting the new drivers and compatibles Cuda and Cudnn versions.