stan-cmdstanr-gpu-docker

Table of Contents

• stan-cmdstanr-gpu-docker
  • Table of Contents
    • Introduction
    • Stan
    • Docker

Introduction

A Docker image to run Stan, cmdstanr, and brms for Bayesian statistical modelling - with GPU support.

Launch an RStudio webserver using bash stan_up.sh.

Execute bash docker_pull.sh to pull the image.

To convert the image into an Apptainer image, run bash apptainer_pull.sh.

Stan

The following packages are installed during the image build process:

• Stan
• rstan
• rstanarm
• cmdstan
• brms
• posterior
• tidybayes
• bayesplot
• posterior
• rstantools
• projpred
• shinystan

The brms_within_chain_parallelization.R script can be executed within the Docker container to evaluate whether within-chain parallelization, CmdStan, brms, and OpenCL are working properly.

Docker

This Docker image extends from rocker/tidyverse. Click this link for more information about the Rocker project.

Running the Docker container will launch an RStudio web server. You can access RStudio by visiting localhost:$R_STUDIO_PORT on your web browser. See .env for the defined environment variables.

Running docker-compose will bind a volume, mapping the container's home directory to a local r_home directory.

See the Dockerfile for the instructions executed during the build of the Docker image.

View docker-compose.yaml to see the definition for the Stan service.