AI Labs Neuroevolution Algorithms
This repo contains distributed implementations of the algorithms described in:
Our code is based off of code from OpenAI, who we thank. The original code and related paper from OpenAI can be found here. The repo has been modified to run both ES and our algorithms, including our Deep Genetic Algorithm (DeepGA) locally and on AWS.
Note: The Humanoid experiment depends on Mujoco. Please provide your own Mujoco license and binary
The article describing these papers can be found here
Visual Inspector for NeuroEvolution (VINE)
./visual_inspector contains implementations of VINE, i.e., Visual Inspector for NeuroEvolution, an interactive data visualization tool for neuroevolution. Refer to
README.md in that folder for further instructions on running and customizing your visualization. An article describing this visualization tool can be found here.
Accelerated Deep Neurevolution
./gpu_implementation contains an implementation that uses GPU more efficiently. Refer to
README.md in that folder for further instructions.
How to run locally
git clone https://github.com/uber-common/deep-neuroevolution.git
create python3 virtual env
python3 -m venv env . env/bin/activate
pip install -r requirements.txt
If you plan to use the mujoco env, make sure to follow mujoco-py's readme about how to install mujoco correctly
launch sample ES experiment
. scripts/local_run_exp.sh es configurations/frostbite_es.json # For the Atari game Frostbite . scripts/local_run_exp.sh es configurations/humanoid.json # For the MuJoCo Humanoid-v1 environment
launch sample NS-ES experiment
. scripts/local_run_exp.sh ns-es configurations/frostbite_nses.json . scripts/local_run_exp.sh ns-es configurations/humanoid_nses.json
launch sample NSR-ES experiment
. scripts/local_run_exp.sh nsr-es configurations/frostbite_nsres.json . scripts/local_run_exp.sh nsr-es configurations/humanoid_nsres.json
launch sample GA experiment
. scripts/local_run_exp.sh ga configurations/frostbite_ga.json # For the Atari game Frostbite
launch sample Random Search experiment
. scripts/local_run_exp.sh rs configurations/frostbite_ga.json # For the Atari game Frostbite
visualize results by running a policy file
python -m scripts.viz 'FrostbiteNoFrameskip-v4' <YOUR_H5_FILE> python -m scripts.viz 'Humanoid-v1' <YOUR_H5_FILE>
The extra folder holds the XML specification file for the Humanoid Locomotion with Deceptive Trap domain used in https://arxiv.org/abs/1712.06560. Use this XML file in gym to recreate the environment.
How to run in docker container
You can also run the code inside a docker container using docker and docker-compose.
Clone repo and enter the directory.
git clone https://github.com/uber-common/deep-neuroevolution.git cd deep-neuroevolution
Start the container launching the redis instance, use sudo if required, see also this page.
sudo docker-compose up
Open up a second terminal session into the container.
sudo docker exec -it deepneuro /bin/bash
Start the experiment of your choice as stated above. E.g.
cd ~/deep-neuroevolution/ . scripts/local_run_exp.sh es configurations/frostbite_es.json