keras does not pick up tensorflow-gpu

I believe I have all the right stuff installed. When I do 'install_keras(tensorflow="gpu)' it appears to run without error. However, back in the R script, at 'library(keras)' i get an error 'Tensorflow not found.'

Runs fine with cpu. Has run fine before I reloaded the lubuntu 18.04 OS. In the record of gpu installation, the computer records picking up tensorflow 1.13.1.

I think some file(s) are in the wrong place but just can't figure it out. Any help will be appreciated.

Best regards

Mike Gallagher

1 Like

Usually this behavior is related to GPU drivers misconfigurations. What's the output of running nvidia-smi on your terminal?

mike@mike-asrock:~$ nvidia-smi
Fri Apr 26 15:41:16 2019
| NVIDIA-SMI 390.116 Driver Version: 390.116 |
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| 0 GeForce GTX 106... Off | 00000000:01:00.0 On | N/A |
| 0% 50C P8 6W / 120W | 36MiB / 3015MiB | 0% Default |

| Processes: GPU Memory |
| GPU PID Type Process name Usage |
| 0 732 G /usr/lib/xorg/Xorg 34MiB |

It seems from the output that you don't have CUDA installed, and this would cause tensorflow-gpu to not load correctly.

For example, in the output should print what version of cuda you have installed


Looks like I am going to have to start over.

In the meantime, though, is this url information still the latest available? I've used it before.


It's always hard to install the NVIDIA requirements.
I try to follow the steps here very closely.

1 Like

Thanks. I was going to come around to asking if there were not a more recent script for the GPU install than the one I had.

BTW I have managed to work through putting together NVIDIA 410, CUDA 9.0 and the latest CUDnn for 9.0. So we did manage to solve my problem!

Best regards

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.