Installation#

In order to successfully install LabGym, you will need to have the following dependencies installed and available on PATH:

  • Python 3.9 or 3.10

  • Git

  • A C/C++ compiler (e.g. gcc, clang, or Microsoft Visual C++)

If you are on a Windows or Linux system with an Nvidia GPU, you will need the following installed to enable GPU-accelerated training:

  • CUDA Toolkit 11.8

  • cuDNN

We recommend that you follow the instructions below for your operating system to ensure that the installation goes smoothly. However, if your system configuration requires that you install these dependencies in a different way, (e.g. you’re using a computing cluster that uses Lmod to load Python) feel free to go ahead and do that.

Regardless of your system configuration, we strongly recommend using pipx to install LabGym. If you use Python for other research tasks, you might have multiple version of Python installed, as well as versions of libraries like NumPy and TensorFlow that LabGym is incompatible with. Using pipx ensures that LabGym and its dependencies can be installed in an isolated environment with the correct version of Python, while allowing the LabGym command to be recognized globally in the terminal. For more information, please see the system-specific instructions below.