Installation#

Installing the Latest Release#

SOCKS requires python 3.7 or greater to work.

The toolbox currently supports linux and macOS systems, but we do not officially support Windows. The code should still work on Windows systems, but this is currently untested.

In order to avoid conflicts with other packages, it is highly recommended to use a virtual environment.

Install python3 using the package manager.

For instance, on Ubuntu 20.04, ensure that python3 is installed on your system using python3 --version, and make sure that the version is greater than 3.7. Then, to install pip, use:

apt update & apt install python3-pip

Install python3 via homebrew brew install python or from http://python.org.

Install python3, for instance from http://python.org.

Then, run the following command:

pip install gym-socks

Alternatively, to install from source, download the latest code from GitHub, and from the code directory, run the following command:

pip install .

A Dockerfile is provided with the repo which installs all necessary dependencies and installs SOCKS automatically. First, install Docker on your system. Then, to build the docker image from the Dockerfile, use the following command:

docker build -t socks .

Then, to launch the socks docker container, use:

docker run -it socks bash

We use this docker container for all development and testing.

Troubleshooting#

If you are having issues installing the code, open an issue at https://github.com/ajthor/socks/issues and describe your issue, including all relevant error messages.

Common Issues#

  • SOCKS currently depends on scikit-learn, numpy, and scipy to work. In our experience, installing these packages via linux package managers can lead to difficult to diagnose errors in the package setup. For instance, on Ubuntu, we recommend installing libblas-dev and liblapack-dev using apt, and then installing scikit-learn, numpy, and scipy via pip.

    apt update & apt install libblas-dev liblapack-dev
pip install numpy scipy scikit-learn gym-socks