This repository is the official implementation of Learning State Representations from Random Deep Action-conditional Predictions.

Run the following command to create a new conda environment with all dependencies:

conda env create -f conda_cuda11_cp38_linux64.yml

Then activate the conda environment by

conda activate rgvfs

Or if you prefer using your own Python environment, run the following command to install the dependencies:

pip install -r requirements.txt

OpenAI Gym does not contain the ROMs for the Atari games. Please refer to atari-py for how to download and import the ROMs.

Please refer to the instructions at https://github.com/deepmind/lab for installation.

Run the following command for the Atari experiment:

python -m experiments.atari

Run the following command for the DeepMind Lab experiments:

python -m experiments.dmlab

To run the stop-gradient experiments, use the argument --stop_gradient. For example, the following command runs the stop-gradient experiment in Atari

python -m experiments.atari --stop_gradient