Skip to content

Instantly share code, notes, and snippets.

@seominjoon
Last active October 20, 2020 03:33
Show Gist options
  • Save seominjoon/1f461cdd8e17c2a236c9 to your computer and use it in GitHub Desktop.
Save seominjoon/1f461cdd8e17c2a236c9 to your computer and use it in GitHub Desktop.
Installing torch with GPU enabled on AWS EC2 (g2 instance) Ubuntu 14.04.
# install torch on ec2 g2 instance
# installing torch
curl -s https://raw.githubusercontent.com/torch/ezinstall/master/install-deps | bash # prereqs
git clone https://github.com/torch/distro.git ~/torch --recursive
cd ~/torch; ./install.sh
source ~/.bashrc
# confirm torch is installed
th
# update linux headers
# for some reason, if these are not installed, then Nvidia drivers will not work
sudo apt-get update
sudo apt-get install -y linux-source linux-image-extra-`uname -r` linux-headers-`uname -r` linux-image-`uname -r`
# install nvidia driver and CUDA
sudo apt-get install -y cuda
# check if you have gpu (or 4 gpus) working:
nvidia-smi -l
lsmod | grep -i nvidia
# obtain cudnn. For torch, you need v3 (v4 doesn't work).
# Note that these are not public; you have to register and download them yourself (takes 1-2 days)
wget https://YOUR_URL/cudnn-7.0-linux-x64-v3.0-prod.tgz
tar -zxvf cudnn-7.0-linux-x64-v3.0-prod.tgz
# move the files to /usr/local/cuda
# Install all different packages
sudo apt-get install -y libprotobuf-dev # needed for loadcaffe
luarocks install cutorch # CUDA support for torch
luarocks install cunn # CUDA neural net?
luarocks install hdf5 # for processing .h5 files.
luarocks install cudnn # cudnn support for torch
luarocks install loadcaffe
# and other stuff..
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment