Skip to content

Latest commit

 

History

History
30 lines (18 loc) · 964 Bytes

Torch.rst

File metadata and controls

30 lines (18 loc) · 964 Bytes

Torch on ShARC

Torch is a scientific computing framework with wide support for machine learning algorithms that puts GPUs first. It is easy to use and efficient, thanks to an easy and fast scripting language, LuaJIT, and an underlying C/CUDA implementation.

The following is an instruction on how to setup Torch on your local user account.

Request an interactive session e.g.

qrshx -l gpu=1

Load the relevant modules for ShARC

module load apps/python/anaconda3-4.2.0
module load libs/cudnn/5.1/binary-cuda-8.0.44

Create a conda environment to load relevant modules on your local user account and activate it

conda create -n torch python=3.5 anaconda3-2.5.0
source activate torch

Install CMake, this is needed to build the Torch binaries

conda install cmake

Download Torch distro from github and run the install script

git clone https://github.com/torch/distro.git ~/torch --recursive
cd ~/torch
./install.sh