CGI DRL Platform

This repository is the official release of DRL Platform of CGI Lab (https://cgi.lab.nycu.edu.tw/), Department of Computer Science, National Yang Ming Chiao Tung University. We implement some of popular deep reinforcement learning algorithm on some simulator or environemnts, especially, in video games. Some of algorithms proposed by our lab will also be put at this repo in a specific branch:

Publication Implementations

• Chiu-Chou Lin, Wei-Chen Chiu, I-Chen Wu. An unsupervised video game playstyle metric via state discretization. UAI 2021. Link branch Dataset and HSD Models
• Kuo-Hao Ho, Ping-Chun Hsieh, Chiu-Chou Lin, You-Ren Lou, Feng-Jian Wang, I-Chen Wu. Towards Human-Like RL: Taming Non-Naturalistic Behavior in Deep RL via Adaptive Behavioral Costs in 3D Games. ACML 2023. Link branch
• Chiu-Chou Lin, Wei-Chen Chiu, I-Chen Wu. Perceptual Similarity for Measuring Decision-Making Style and Policy Diversity in Games. TMLR 08/2024. Link branch New Dataset
  • This is a backup code repository (the version after the last paper revision). If you have any problems running this code, feel free to create a new issue in this repo.
• Chiu-Chou Lin, Yu-Wei Shih, Kuei-Ting Kuo, Yu-Cheng Chen, Chien-Hua Chen, Wei-Chen Chiu, I-Chen Wu. Identifying and Clustering Counter Relationships of Team Compositions in PvP Games for Efficient Balance Analysis. TMLR 09/2024. Link branch dataset and models
  • This is a backup code repository (the version after the last paper revision). If you have any problems running this code, feel free to create a new issue in this repo.

Environment Installation

1. Get CGI DRL Platform

cd ~
git clone https://github.com/DSobscure/cgi_drl_platform.git

2. Install Podman official instructions (if you cannot use GPU with podman on some machines, e.g., AWS, you can use docker)

# only for ubuntu
. /etc/os-release
echo "deb https://download.opensuse.org/repositories/devel:/kubic:/libcontainers:/stable/xUbuntu_${VERSION_ID}/ /" | sudo tee /etc/apt/sources.list.d/devel:kubic:libcontainers:stable.list
curl -L "https://download.opensuse.org/repositories/devel:/kubic:/libcontainers:/stable/xUbuntu_${VERSION_ID}/Release.key" | sudo apt-key add -
sudo apt-get update
sudo apt-get -y upgrade
sudo apt-get -y install podman

3. Create Container

Build local dockerfile (use a propoer CUDA driver version on the server, we use CUDA11.6 and NVIDIA built TF1.15 version as the base file)

podman build -t cgi_drl -f $PWD/cgi_drl_platform/infrastructure/dockerfiles/CUDA11.6/Dockerfile .

create a podman container

podman run -d -v $PWD:/root -w /root --name cgi_drl -it cgi_drl bash

4. Enter Container and Install cgi_drl_platform

podman exec -it cgi_drl bash
cd cgi_drl_platform/platform
pip install -e .

5. Download Dataset and HSD Models

Dataset and HSD Models

• upzip playstyle_uai2021.zip at /root/ in the container
• New Dataset and HSD in TMLR 08/2024 will be updated in the future
  • For Go experiments, the dataset will not be uploaed due to the privacy issue. New Dataset
• upzip playstyle_similarity_tmlr.zip at /root/ in the container
  • This dataset including 2048 playing results amd DRL diversity testing with first IQN model.
  • For Go experiments, the dataset will not be uploaed due to the privacy issue. If you are interesting in that part, please contact the MiniZero team.

Run Experiemnts

cd ~/cgi_drl_platform/platform/cgi_drl
python run.py -k rgsk-metric-24style