Right, yesterday during one of the reading groups meeting we decided to have the next session on TensorFlow installation on DICE machines. So I decided to test it out and find why it is considered to be problamatic.
Turns out its very simple!
I’d recommend that you use VirtualEnv as otherwise you’d need
sudo access to the machine.
Right so here are the steps for installing the GPU version:
- Activate your VirtualEnv
pip install --upgrade https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow-0.7.1-cp27-none-linux_x86_64.whl
- Check the version of
cudaon your machine. On DICE its in the
export PATH=$PATH:"/opt/cuda-X.X.XX/bin"where its cuda-7.5.18 for my machine
Equivalently you can append these things to your config file.
Thats it!! You are good to go.