Install GPI

This software is made available for download via this website subject to the license terms and conditions, available also on this website. You will be required to view and accept these license terms and conditions upon installation of the software. If you do not accept or agree to these license terms, you are not permitted to download, install or otherwise use the software in any way. By continuing with this download, you are agreeing to the applicable license terms and conditions.

Minimum Hardware

It is recommended to use at least a dual processor system with 4GB of system memory for GPI.


Read This First

GPI is built and distributed on Conda-Forge, a community for open-source software distributed using the package manager conda. Conda allows you to create and maintain many separate "environments", each containing different software (such as different versions of Python). You can also save snapshots of an environment to help others verify or reproduce your work using exactly the same software versions.

The most common conda distribution is Anaconda, which contains the package manager, a number of commonly used packages, and a graphical interface that may be easier for new GPI users. If you'd like to install GPI using Anaconda and the GUI (Anaconda Navigator), follow the guide for installing with Anaconda Navigator.

GPI can also be installed, configured, and launched directly from the command line. This install method will be more appropriate for advanced users or those who are already using conda from the command line. If you prefer this method, skip ahead to the instructions for installing from the command line.

Installing with Anaconda Navigator

  1. If you already have Anaconda installed, go ahead to step 2. If not, download and run the latest Anaconda installer here (choose the "Python 3.7" version). When prompted, make sure to respond "yes" to the question about initializing Anaconda3 with conda init.

    Advanced notes (click to expand)
    The installer will typically modify ~/.bashrc or ~/.bash_profile (or ~/.zshrc, on OSX 10.15) so that new terminal windows open with the base conda environment active. If you'd prefer terminal windows to open outside of any conda environment, run the command:

    conda config --set auto_activate_base false

    The installer may also ask if it should add conda to your PATH. We recommend responding "no" (the default) to avoid breaking programs that rely on the system Python. For more information on Anaconda installation options, see the Anaconda FAQ.

  2. If Anaconda Navigator is not already open, launch it as follows:

    • Mac - Choose the icon in your Applications folder.
    • Windows - Search for "Anaconda Navigator" in the Start Menu.
    • Linux - Open a terminal window and enter the command anaconda-navigator (note that if you set auto_activate_base to false in step 1, you'll have to enter conda activate base first - and you may find it more straightforward to use the command line installation method below).
  3. In Navigator, select Environments at left, then click Create at the bottom of the screen. Enter a name for this new environment (we'll call it gpi_env for the rest of these instructions). Make sure the box next to Python is checked, and select version 3.7. Click Create. The window should disappear, and after a few seconds you should see gpi_env show up in the environment list.

  4. There should be a triangle to the right of gpi_env indicating that it is the active environment. If this isn't the case, first double-click on gpi_env to activate it. Then click the triangle and choose Open Terminal to bring up a command prompt.

  5. A terminal window should open (on Linux, it may appear silently as a new tab in the terminal from step 2). Enter the following commands:

    conda config --env --add channels conda-forge

    conda config --env --set channel_priority strict

    This prepares your new environment to install the correct packages for GPI, without modifying any other environments.

  6. Close the terminal window/tab launched in step 4. In the right side of the Anaconda Navigator, you should see the list of packages installed in gpi_env. Above that list, click Update Index to refresh the list of available packages.

  7. Once this has finished running, change Installed to Not Installed in the drop-down menu, then type gpi into the search box at right. You should see a list of packages including gpi and gpi_core. You may also see other public GPI node libraries, which will have names starting with gpi_. Select gpi and gpi_core (and others, if you want to install them), then click Apply.

  8. A pop-up window will open and display a list of packages that will be added to gpi_env. You can take a look through this list, if you're interested in what software GPI depends on behind the scenes. When you're ready, click Apply again. The installation may take several minutes.

  9. Run GPI: Once installation is complete, you should see GPI listed under the Home tab, with a Launch button available. You can run GPI using this button (in our experience, the first launch may be a bit slow). Each time you close and re-open Anaconda Navigator, you'll need to select gpi_env from the drop-down menu or Environments tab for GPI to appear.

    Advanced notes (click to expand) You may also wish to launch GPI by opening the gpi_env terminal (as in step 4) and entering the command gpi. This will allow you to see GPI warnings and error messages, which are currently not visible when launching from the Home tab (we're working on a fix!). Launching a gpi_env terminal also allows you to use the gpi_make command for building C++ extensions with PyFI.

Installing from the command line

  1. If you already have a conda distribution installed, skip ahead to step 2. If not, download and install Anaconda here, or Miniconda (a lighter version about 90% smaller than Anaconda) here. When prompted during installation, answer "yes" to the question about initializing conda using conda init. We recommend the default response of "no" when asked if you want to add conda to your system PATH, as this may break other things that depend on your system's installation of Python. For more information, please see the Anaconda FAQ.

    Note If you do not want terminal windows to open with the "base" conda environment active by default, you can disable this behavior with the following:

    conda config --set auto_activate_base false

  2. Launch a new terminal instance. Windows users with Anaconda can launch "Anaconda Prompt" from the Start Menu, or any other terminal (cmd.exe, PowerShell, or MinGW) for which conda has been configured. Next, choose a name for the environment that will hold GPI. We will use gpi_env for this tutorial. Finally, run the following commands to set up and configure the new environment:

    conda create -n gpi_env

    conda activate gpi_env

    conda config --env --add channels conda-forge

    conda config --env --set channel_priority strict

    Advanced notes (click to expand) These commands assume you are starting from a clean installation of conda. If you have made config changes to conda prior to this installation, you may need to pay attention to those settings. The end goal is to have a new environment with at least two channels, conda-forge (first) and defaults (second), and with the channel_priority attribute set to "strict". You can omit the --env flag in the above commands if you want these configuration changes to apply to all of your conda environments. See here for a full description of conda configuration options.

  3. Install GPI and the core nodes. In the same terminal as above (making sure gpi_env is still the active environment), run the following command:

    conda install gpi_core python=3.7 pyqt=5.9

    Python 3.7 and Qt 5.9 currently provide the "smoothest" running GPI, but you can also install with Python 3.6 or 3.8, and Qt 5.6 or 5.12 (the latter is required for Python 3.8). If you omit the python and pyqt pinnings from the above command, conda will use the newest possible version from conda-forge for both packages.

  4. Running GPI - GPI is now installed and ready to use. Run it with the command gpi. When starting from a new terminal, remember that you will first need to activate the environment gpi_env using conda activate gpi_env.

Updating GPI

You can update GPI like any other conda package, either from within Anaconda Navigator or using the command line. In addition, GPI contains a built-in updater, which is found under: - GPI > Search For Updates (Mac) - Help > Search For Updates (Windows/Linux)

Additional Notes

If you are upgrading from a very early version of GPI, v0.5 or older, first move (or remove) the existing installation, which will most likely be located at /opt/gpi.

If you are a Mac user upgrading from the previous App-packaged version, simply delete the older GPI in your Applications folder prior to installing the new one. You may also need to modify the file ~/.bashrc or ~/.bash_profile to remove links to the App-style GPI package (which included a hidden installation of Miniconda).

MacPorts Installation

Our thanks to Eric Borisch for enabling GPI installation via MacPorts. To install via MacPorts, use the following commands:

$ sudo port install py-gpilab-framework
$ sudo port install py-gpilab-core

Other Notes

  • If you are having issues with node libraries not being visible, make sure to check your ~/.gpirc file for the correct library path. By default, the packaged gpi_core library and any library under ~/gpi is searched.