Managing python versions in RStudio

Python
Published

July 5, 2023

Ressources

  • Good resources are found in the {reticulate} vignettes: Link, in particular this
  • Link to question on stackoverflow - Link
  • Quarto documentation explaining how to use different environments: Link
  • An interesting blog post how to managage package versions in python: Link

Managing python installation

library('reticulate')
py_discover_config()
py_config()
conda_list()

conda_version()
conda_create('sbloggel')


conda_install('sbloggel', 'pandas')
miniconda_update()
use_condaenv('sbloggel')
conda_install('sbloggel', 'plotly')
use_condaenv('base')

Install virtual environments to jupyter

# Prepare python to enable installation of virtual environments in Jupyter:
pip install --user ipykernel

# Show conda environments and activate:
conda info --envs  
conda activate sbloggel

# Install virtual environment into jupyter:
python -m ipykernel install --user --name=sbloggel
conda activate /Users/seb/Library/r-miniconda-arm64/envs/sbloggel

Start Jupyter from terminal:

jupyter notebook

Installing tensorflow_decision_forests

  • I couldn’t install it via conda, so I created a new environment and installed it via pip (this seems also to be the recommended way on the tensorflow website):
library('reticulate')
conda_create('tensorflow', python_version = '3.11')
use_condaenv('tensorflow')
conda_install('tensorflow', c('tensorflow', 'tensorflow_decision_forests'), pip = TRUE)
conda_install('tensorflow', c('pandas', 'matplotlib', 'seaborn'), pip = TRUE)
conda_install('tensorflow', 'IPython', pip = TRUE)
py_version()

Installing environment - rebuild sbloggel

library('reticulate')
conda_create('sbloggel', python_version = '3.11')
packages_to_install <- c('pandas', 'matplotlib', 'seaborn', 'plotly', 'statsmodels', 'scipy', 'scikit-learn', 'pytorch', 'torchvision')
conda_install('sbloggel', packages_to_install)
use_condaenv('sbloggel')
py_version()
10 + 10
import pandas as pd
import fastai as fa
import torch
device = (
    "cuda"
    if torch.cuda.is_available()
    else "mps"
    if torch.backends.mps.is_available()
    else "cpu"
)
print(f"Using {device} device")