GPU is not utilized

1

I am implementing the following project: https://github.com/sinhasam/vaal
and it is not using the system’s GPU. Kindly check main.py

  • PyTorch Version : 1.1.0
  • OS: Ubuntu
  • How you installed PyTorch: pip
  • Python version: 3
  • CUDA/cuDNN version:
  • GPU models and configuration:
    nvcc: NVIDIA ® Cuda compiler driver
    Copyright © 2005-2016 NVIDIA Corporation
    Built on Tue_Jan_10_13:22:03_CST_2017
    Cuda compilation tools, release 8.0, V8.0.61
2

How did you install PyTorch?
Did you use the binaries with the cudatoolkit?
If so, could you post the output of print(torch.cuda.is_available())?

3

I am running the same program on server as well as on my system.

At my system Pytorch is installed by pip. I dont know about server’s installation. But at both systems

torch.cuda.is_available()
True
Device Count at server is 4 and at my system it is 1
torch.cuda.device_count()
4

No, I didn’t use cudatoolkit

Screenshot from 2020-02-17 13-32-17
Screenshot from 2020-02-17 13-32-17956×588 81.2 KB

4

Thanks for the update.
Does the device memory increase during the execution of the script and how does “GPU-Util” print?

5

No, it is not increasing thats way I am sure that program is not utilizing gpus.

6

I assume you’ve passed the --cuda argument when running main.py?

7

No I didnt because there was no such instruction for running the main.py

I am implementing this project from github https://github.com/sinhasam/vaal

8

This seems to be necessary as seen in these lines of code.

1 Like
9

OK I will try again.
Thank you!