Keras/Tensorflow Running slowly on Windows 10

Accelerated Computing CUDA CUDA Programming and Performance
1

Operating system: Windows 10 Professional
Python: 3.6.5
Tensorflow: 1.13.0rc2
Cuda: V10.0.130
Cudnn: cudnn-10.0-windows10-x64-v7.4.2.24

The training for a LSTM model is ~5X slower than that on Linux system. How cam I speed it up? Thank you.

2

Forgot to mention that the hardware is a RTX 2080 ti with 11GB memory.

3

TF 1.13.1 was released Monday with support for CUDA 10. You might want to try it instead of 1.13.0rc2

And you might want to file a TF issue if you’re having trouble.

4

I upgraded TF to 1.13.1 as you suggested. The performance is still the same.

Should I upgrade Cudnn to the newest version as well?

Thank you.

5

Hi, Robert,

Now I have Ubuntu on Win10. And installed driver version 418 by

$ sudo apt install nvidia-driver-418

It ran through just fine. But

$ ubuntu-drivers list

is empty and

$ nvidia-smi
NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Please help me, thank you.

6

I have rebooted the system after installation.

7

Ubuntu 18.04.2 LTS

8

You can’t install ubuntu on windows and use the GPU in that setting.

9

Ok, thanks, Robert.