choose GPU device

[kingfengji]

hi there, great one!

one question, I got 3 GPUs to work with, how to choose which GPU to use?

I mean, I dont want to use gpu_0 cause my monitors are connected with it.
I wish to use GPU_1 instead...

THanks!

[anishathalye]

There's currently no clean way to do it via command line switches, etc - you need to modify the code.

See how we specify that we want to use the CPU here? You can do something similar here and force it to use /gpu:1. See this TensorFlow documentation for more information.

[anishathalye]

Btw, if you want to improve the code so that you can do this kind of stuff via command line switches, pull requests are welcome 😄

[rayrayson]

Have you tried setting the "CUDA_VISIBLE_DEVICES" environment variable before starting neural_style.py ?

Ref CUDA doc: http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#env-vars

It worked for us (the Open Grid Scheduler/Grid Engine project) when we needed to control which GPU is used in a job when we did the Multi-Core Processor Binding with hwloc project:

http://gridscheduler.sourceforge.net/projects/hwloc/GridEnginehwloc.html

[anishathalye]

Going to assume this is resolved. If not, feel free to reopen.