2. Installation

The pysdg library can be used in both python or R. The package can be installed either using wheel file or Docker images.

The Docker images for Python are available at pysdg Python images. while these for R are available at pysdg R images. The images are tagged with the version of the package.

Follow the steps provided in the readme file to install the package using the Docker images.

Below are the details to install the library using the wheel files.

2.1. Python

1. From your terminal, create your virtual environment using conda and make sure to specify the python version as 3.10.14 and the pip version as 24.0, e.g.

   conda create -n  <your_virtual_environment_name> python=3.10.14
   conda activate <your_virtual_environment_name>
   pip install pip==24.0
   

Tip

If you are using RUNAI, it is advisable to create your Conda environment outside the pod, such as in your personal folder. This avoids the need to recreate the environment each time the pod is terminated. You can do this by specifying the –prefix option when creating the environment. For example, in any terminal, run:

conda create –prefix /share/personal/your_username/your_env_dir/your_env_name python=3.10.14

Afterward, when starting a new pod, you can direct Conda to look for environments in that folder by running:

conda config –add envs_dirs /share/personal/username/your_env_dir

Finally, activate the environment using:

conda activate your_env_name

2. Download the required release of pysdg wheel file from pysdg wheel file releases.

3. If needed, and if you have the necessary credentials, download the wheel file provided to you by Replica/Aetion. You do not need to download this file if you are not using the Replica generator.

4. Install pysdg and Replica (if needed) using:

   pip install <the_downloaded_pysdg_wheel_file>
   pip install <the_downloaded_replica_wheel_file>
   

5. If you intend to use the package in Jupyter Notebook, you need to add the above environment as a Jupyter kernel by issuing the following additional commands from your terminal:

   conda install -c anaconda ipykernel
   python -m ipykernel install --user --name=<your_virtual_environment_name>
   

Then from your Jupyter notebook, select the new kernel that was created based on your environment.

2.2. R

From your R console, execute teh following steps:

1. Install the package reticulate and the python version 3.10.14 through R. (Note: Installing python using reticulate may take a long time):

   devtools::install_version("reticulate", version = "1.38.0")
   reticulate::install_python(version="3.10.14")
   

   Tip

   Typically, if you are using Rstudio template when creating a runai job, installing Python should execute smoothly. However, in case you encounter any errors, then, from your RStudio terminal, try updating Linux using:

   sudo apt-get update
   sudo apt-get install -y build-essential libssl-dev zlib1g-dev libncurses5-dev libgdbm-dev libnss3-dev libsqlite3-dev libreadline-dev libffi-dev libbz2-dev liblzma-dev
   

2. Install miniconda to create a conda environment:

   reticulate::install_miniconda()
   reticulate::conda_create("<your_virtual_environment_name>", python_version = "3.10.14")
   

3. Make sure you use the correct version of pip:

   reticulate::use_condaenv("<your_virtual_environment_name>")
   reticulate::conda_install("<your_virtual_environment_name>", packages = "pip==24.0", pip = FALSE)
   

4. Follow steps 2 and 3 under Python installation above.

5. Install pysdg within the conda environment:

   reticulate::conda_install("<your_virtual_environment_name>", "path_to_the_downloaded_pysdg_wheel_file", pip = TRUE, conda = "auto", python_version = "3.10.14")
   

6. Use the same environment above to install Replica (if needed) using:

   reticulate::conda_install("<your_virtual_environment_name>", "path_to_the_downloaded_replica_wheel_file", pip = TRUE, conda = "auto", python_version = "3.10.14")