0

Can I run the DeepLab image segmentation completely on CPU?

I have access to hpc, with high Memory resources but it is not GPU-enabled.

behnam
  • 25
  • 4

2 Answers2

1

yes you can run it completely on the CPU. For that you only have to make a small change:

Open the file train.py and include the line

os.environ["CUDA_VISIBLE_DEVICES"]=""

before tensorflow is included.

FranklynJey
  • 646
  • 4
  • 8
  • Thank you! This worked. I got access to a GPU enabled cluster, though. It works much faster now. I cannot set batch size more than 12, but it's still fast – behnam Nov 21 '19 at 22:07
0

Yes, you can. In fact you might not need to change the code at all to run it on a CPU-only machine.

hkchengrex
  • 4,361
  • 23
  • 33