I am trying to run DL PyTorch jobs on a GCP box. Steps I take:
- Create a VM with m36 deep learning image and K80 attached. There is a question re CUDA 10 driver installation at first login, I choose 'y' and the driver gets set up.
- install pyenv
pyenv install 3.7.5pyenv virtualenv 3.7.5 myenvpyenv activate myenvpip install fastai- Start python console and run:
Code:
import torch
torch.cuda.is_available()
It returns False. What am I supposed to do to enable GPU in PyTorch?