sjSDM: Help with the installation of dependencies

Maximilian Pichler & Florian Hartig, Theoretical Ecology, University of Regensburg

2022-06-22

Abstract

This document explains how to install dependencies for the sjSDM package.

The r sjSDM::install_sjSDM() function can install automatically all necessary ‘python’ dependencies but it can fail sometimes because of individual system settings or if other ‘python’/‘conda’ installations get into the way.

PyTorch Installation - Before you start:

A few notes before you start with the installation (skip this point if you do not know conda):

Windows - automatic installation:

Sometimes the automatic ‘miniconda’ installation (via r sjSDM::install_sjSDM() ).doesn’t work because of white spaces in the user’s name. But you can easily download and install ‘conda’ on your own:

Download and install the latest ‘conda’ version

Afterwards run:

install_sjSDM(version = c("gpu")) # or "cpu" if you do not have a proper gpu device

Reload the package and run the example, if this doesn’t work:

Windows - manual installation:

Download and install the latest ‘conda’ version

Open the command window (cmd.exe - hit windows key + r and write cmd)

Run in cmd.exe:

conda create --name r-sjsdm python=3.7
conda activate r-sjsdm
conda install pytorch torchvision cpuonly -c pytorch # cpu
conda install pytorch torchvision cudatoolkit=11.3 -c pytorch #gpu
python -m pip install pyro-ppl torch_optimizer madgrad

Restart R, try to run the example, and if this doesn’t work:

Linux - automatic installation:

Run in R:

install_sjSDM(version = c("gpu")) # or "cpu" if you do not have a proper gpu device

Restart R try to run the example, if this doesn’t work:

Linux - manual installation:

We strongly advise to use a ‘conda’ environment but a virtual environment should also work. The only requirement is that it is named ‘r-sjsdm’

Download and install the latest ‘conda’ version

Open your terminal and run:

conda create --name r-sjsdm python=3.7
conda activate r-sjsdm
conda install pytorch torchvision cpuonly -c pytorch # cpu
conda install pytorch torchvision cudatoolkit=11.3 -c pytorch #gpu
python -m pip install pyro-ppl torch_optimizer madgrad

Restart R try to run the example, if this doesn’t work:

MacOS - automatic installation:

Run in R:

install_sjSDM()

Restart R try to run the example, if this doesn’t work:

MacOS - manual installation:

We strongly advise to use a ‘conda’ environment but a virtual environment should also work. The only requirement is that it is named ‘r-sjsdm’

Download and install the latest conda conda version

Open your terminal and run:

conda create --name r-sjsdm python=3.7
conda activate r-sjsdm
python -m pip install torch torchvision torchaudio 
python -m pip install pyro-ppl torch_optimizer madgrad

Restart R try to run the example, if this doesn’t work:

Help and bugs

Example

library(sjSDM)
community <- simulate_SDM(sites = 100, species = 10, env = 5)
Env <- community$env_weights
Occ <- community$response

model <- sjSDM(Y = Occ, env = linear(Env, ~0+X1*X2 + X3 + X4))
summary(model)