Skip to content

Installation of supported environments#

Below we provide installation instructions for a number of environments that we support.

Installation of MiniGrid#

MiniGrid will automatically be installed when installing allenact and so nothing additional needs to be done. If you wish to (re)install it manually using pip, simply run the command:

pip install gym-minigrid

Installation of iTHOR#

iTHOR will automatically be installed when installing allenact and so, if you have installed allenact on a machine with an attached display, nothing additional needs to be done. If you wish to (re)install it manually using pip, simply run the command:

pip install ai2thor

The first time you will run an experiment with iTHOR (or any script that uses ai2thor) the library will download all of the assets it requires to render the scenes automatically.

Trying to use iTHOR on a machine without an attached display?

If you wish to run iTHOR on a machine without an attached display (for instance, a remote server such as an AWS machine) you will also need to run a script that launches xserver processes on your GPUs. This can be done with the following command:

sudo python scripts/startx.py &

Notice that you need to run the command with sudo (i.e. administrator privileges). If you do not have sudo access (for example if you are running this on a shared university machine) you can ask your administrator to run it for you. You only need to run it once (as long as you do not turn off your machine).

Installation of RoboTHOR#

RoboTHOR is installed in tandem with iTHOR when installing the ai2thor library. For more information see the above section on installing iTHOR.

Installation of Habitat#

Using Docker#

To run experiments using Habitat please use our docker image using the following command:

docker pull allenact/allenact:latest

This container includes the 0.1.0 release of allenact, the 0.1.5 release of habitat as well as the Gibson point navigation dataset. This dataset consists of a set of start and goal positions provided by habitat. You then need to launch the container and attach into it:

docker run --runtime=nvidia -it allenact/allenact

If you are running the container on a machine without an Nvidia GPU, omit the --runtime=nvidia flag.

Once inside the container simply cd into the allenact directory where all the allenact and habitat code should be stored:

Unfortunately we cannot legally redistribute the Gibson scenes by including them in the above container. Instead you will need to download these yourself by filling out this form and downloading the gibson_habitat_trainval data. Extract the scene assets (.glb files) into habitat-lab/data/scene_datasets/ within the above container. You can then proceed to run your experiments using allenact as you normally would.

Using conda (experimental)#

The following is experimental, we do not guarantee that AllenAct will continue to support this installation procedure in future releases.

Habitat has recently released the option to install their simulator using conda which avoids having to manually build dependencies or use Docker. This does not guarantee that the installation process is completely painless (it is difficult to avoid all possible build issues) but we've found it to be a nice alternative to using Docker. To use this installation option please first install an AllenAct conda environment using the instructions available under the Installing a Conda environment (experimental) section here. After installing this environment, you can then install habitat-sim by running:

If you are on a machine with an attached display:

conda install habitat-sim=0.1.5 -c conda-forge -c aihabitat --name allenact

If you are on a machine without an attached display (e.g. a server):

conda install habitat-sim=0.1.5 headless -c conda-forge -c aihabitat --name allenact