Installing Metaflow


Metaflow is available as an R package for MacOS and Linux. You can get it from our GitHub repository (or very soon from CRAN):

devtools::install_github("Netflix/metaflow", subdir="R")

You can finish the installation by typing -


Metaflow requires Python 3 in order to install the python package since Python 3 has fewer bugs and is better supported than the deprecated Python 2.7. Metaflow R also relies onnumpy and pandasto handle native R matrix and data.frame. The metaflow::install() would install numpy and pandas together with pythonic metaflow package.

If you're inside a conda environment or a Python virtual env, you may want to disable the user mode. See details in the Troubleshooting section below on this topic.


You can test your installation by running:

Sys.setenv("USERNAME" = 'mfuser')

If you see the message Your Metaflow installation looks good! congratulations! Now you can get started with Metaflow by following the tutorial or you can jump straight into the docs.

Now you are ready to get your hands dirty with the Tutorials.

Windows Support

Metaflow currently doesn't offer native support for Windows. However, if you are using Windows 10, then you can use WSL (Windows Subsystem for Linux) to install Metaflow. WSL lets you run a Linux environment inside Windows 10.

Follow these instructions to get set up with WSL 2 -

  1. Update your Windows 10 distribution to version 2004 or higher.

  2. Follow these steps to install WSL 2. When prompted, choose Ubuntu 18.04 as the Linux distribution.

  3. Open Ubuntu 18.04 on your workstation, and run the following commands in the terminal to install R and Python -

sudo apt-key adv --keyserver --recv-keys E298A3A825C0D65DFD57CBB651716619E084DAB9
sudo echo "deb bionic-cran35/" | sudo tee -a /etc/apt/sources.list
sudo apt update
sudo apt install r-base r-base-dev
sudo apt install libcurl4-openssl-dev libxml2-dev libssl-dev
sudo apt-get install python3 python3-pip
echo "alias python=python3; alias pip=pip3" > ~/.bash_aliases
source ~/.bash_aliases
python --version
pip --version
pip install numpy pandas --user

[Optional] To set up RStudio IDE -

  1. Inside Ubuntu 18.04 on Windows, run the following commands -

sudo apt install -y r-base r-base-core r-recommended r-base-dev gdebi-core build-essential libcurl4-gnutls-dev libxml2-dev libssl-dev
sudo gdebi rstudio-server-latest-amd64.deb
sudo rstudio-server start

You can access RStudio at http://localhost:8787 in your browser in Windows.

To install Metaflow -

  1. Inside Ubuntu 18.04 on Windows, open the R console (or your RStudio IDE installed in the previous step) and run the following commands -

install.packages('devtools', INSTALL_opts = c('--no-lock'))
devtools::install_github("netflix/metaflow", subdir = "R")


ERROR: Can not perform a '--user' install

You might see an error message ERROR: Can not perform a '--user' installwhen running metaflow::install(), when your Python interpreter is provisioned by a virtual environment. To unblock yourself, please install by disabling the `user` mode:


Error: Python shared library not found

If you see an error message Error: Python shared library not found when running metaflow::install(), you can set up a conda environment via reticulate and continue your installation inside that conda environment. Here we create a conda environment - metaflow-env :

reticulate::conda_create(envname = "metaflow-env", packages = "python")
reticulate::use_condaenv(condaenv="metaflow-env", required = T)

ModuleNotFoundError: No module named 'metaflow'

If you see the error message ModuleNotFoundError: No module named 'metaflow' right after metaflow::install(), please restart your R session. If this does not help, try installing in a conda environment as mentioned in the previous subsection.

Metaflow could not determine your user name based on environment variables

You need to set the USERNAME environment variable either in bash or in R before running metaflow::test()

export USERNAME=mfuser
Rscript -e "metaflow::test()"
Sys.setenv("USERNAME" = 'mfuser')