This workshop is part of the Rocky Mountain Advanced Computing Consortium’s (RMACC) High Performance Computing Symposium (5/18/2023).

This workshop presents a conceptual overview of the R interface for PyTorch. Emphasis will be on basic structures and applications for deep learning. With torch there is no need to install or use Python. Torch and its associated packages – torchaudio, torchvision, luz – will run on CPU or GPU. This workshop will demonstrate torch for R using an NVIDIA GPU environment, which significantly accelerates computations.

Materials for this workshop are based on the new book, Deep Learning and Scientific Computing with R torch by Sigrid Keydana. For slides click here.

Torch for R is an open source machine learning framework based on PyTorch. torch provides fast array computation with strong GPU acceleration and a neural networks library built on a tape-based autograd system. The "torch for R"" ecosystem is a collection of extensions for torch. For details on getting started, see the Installation Vignette.

1. Install RStudio desktop
2. Install torch

install.packages("torch")

3. Clone this repos and make sure the working directory is torch-workshop
4. Download Tiny ImageNet data

tiny_imagenet_dataset(".", download = TRUE)

5. Install the modeldata package from the 2021-06-08 snapshot

install.packages("modeldata", repos = "https://packagemanager.rstudio.com/cran/__linux__/bionic/2021-06-08")

1. Have a CUDA compatible NVIDIA GPU with compute capability 6.0 or higher
2. Install the NVIDIA Container Toolkit
3. Pull and run the RStudio Rocker container

docker pull rocker/rstudio
docker run --gpus=all -d -t -e PASSWORD=rstudio -p 8787:8787 --name rstudio rocker/rstudio

4. Open RStudio Server from your browser by opening http://<your.ip.address>:8787/. Make sure port 8787 is open. Your username is rstudio and your password is rstudio
5. Configure RStudio Server to download package binaries from the Posit Package Manager

# For Red Hat 8 use:
https://packagemanager.rstudio.com/cran/__linux__/centos8/latest

# For Ubuntu 20.04 use:
https://packagemanager.rstudio.com/cran/__linux__/focal/latest

6. Install torch from the pre-built binaries (Warning! This download is 2Gb)

options(timeout = 600)
install.packages("torch", repos = "https://storage.googleapis.com/torch-lantern-builds/packages/cu117/0.10.0/")

7. Make sure your CUDA device is available (this should return TRUE)

library(torch)
cuda_is_available()

8. Clone this repos and make sure the working directory is torch-workshop
9. Download Tiny ImageNet data

wget http://cs231n.stanford.edu/tiny-imagenet-200.zip
unzip tiny-imagenet-200.zip

10. Install the modeldata package from the 2021-06-08 snapshot

install.packages("modeldata", repos = "https://packagemanager.rstudio.com/cran/__linux__/bionic/2021-06-08")