README.md

In uroshercog/cloudstan-r: Cloudstan R client library

CloudstanR

CloudstanR is an interface to the Cloudstan platform for R users. Note this is a prototype version and there definitely are bugs. Please report them by opening an issue on this repository.

Goals

• No need to install and manage local versions of cmdstan.
• Compilation and sampling on powerful remote executors*.
• Support for high level of paralelization of jobs**.
• Central repository for compiled models, data, and samples.
• Drop-in replacement for cmdstanr.

* Different executor resource configurations are not be available at this time.

** Running more than 4 parallel jobs is not available at this time.

Install

First install the cmdstanr library. This library depends on specific functions from it such as loading results into standard cmdstanr data structures. To install the library please follow the guide here. Note installing the whole cmdstan suite is NOT required if you plan to use only Cloudstan. Otherwise if you wish to toggle between the two (local and remote) install cmdstan as well.

Then you can install the latest version of th cloudstanr library directly from GitHub using a library like Devtools. This will create a folder in your home directory and download a binary which communicates with the Cloudstan platform.

# install.packages("devtools")
devtools::install_github("uroshercog/cloudstan-r")

This library is meant to be a drop-in replacement for the cmdstanr library. If you executed library(cmdstanr) before, execute library(cloudstanr) now.

Next run cloudstanr::install(). This downloads and installs a CLI tool which is used by this R library to communicate with the platform.

Login

To use the platform you must create an account and login. Currently the only supported method is login through Google and GitHub. After executing one of the commands below, a browser window with a page prompting you to sign-in to your e.g. GitHub account will open. Please allow Cloudstan to read basic information about your account (only e-mail). Signing-in will automatically create a new account.

Login with Google

cloudstanr::login_google()

Login with GitHub

cloudstanr::login_github()

Use

Once you've successfully installed and signed-in you can try and follow the official tutorial for compiling and sampling a simple model.

CloudstanR provides a cmdstanr compliant interface. This should allow you to use this library without any modifications to existing codebase.

uroshercog/cloudstan-r documentation built on Dec. 23, 2021, 2:03 p.m.