Installing and Managing Python Environments with reticulate

Overview

The text package relies on Python packages. While textrpp_install() function offers an automated setup, manually installing the dependencies can help with troubleshooting and provides greater control for advanced users.

This guide explains how to install everything step-by-step using the reticulate package in R.

Step 1: Install reticulate

install.packages("reticulate")

Step 2: Create a Conda Environment

Use reticulate to install Miniconda:

reticulate::install_miniconda()

Then create a new conda environment with Python 3.9:

reticulate::conda_create("textrpp_reticulate", packages = "python=3.9")

Step 3: Install Python Packages

Install the required python packages (rpp) needed for the text-package:

rpp_packages <- c(
  "torch==2.2.0",
  "transformers==4.38.0",
  "huggingface_hub==0.20.0",
  "numpy==1.26.0",
  "pandas==2.0.3",
  "nltk==3.8.1",
  "scikit-learn==1.3.0",
  "datasets==2.16.1",
  "evaluate==0.4.0",
  "accelerate==0.26.0",
  "bertopic==0.16.3",
  "jsonschema==4.19.2",
  "sentence-transformers==2.2.2",
  "flair==0.13.0",
  "umap-learn==0.5.6",
  "hdbscan==0.8.33",
  "scipy==1.10.1",
  "aiohappyeyeballs==2.4.4"
)

reticulate::conda_install("textrpp_reticulate", packages = rpp_packages, pip = TRUE)

Step 4: Activate the Environment

To use/initialize the environment in your R session:

reticulate::use_condaenv("textrpp_reticulate", required = TRUE)

You now have a fully working manual installation of the text/textrpp Python environment using reticulate.

Show available conda environments

reticulate::conda_list()

Remove the conda environment

# Restart R
.rs.restartR()  # for RStudio automation


# Now safely remove the environment
# WARNING: this is irreversible
reticulate::conda_remove(envname = "textrpp_reticulate")