An implementation of the DQN algorithm in PyTorch, trained on Super Mario Bros and Atari Pong. The overall architecture is inspired by openai/baselines.
Warning: Training DQN requires sufficient memory. With the default replay buffer size of 1,000,000, it will use about 8 GB of RAM.
- Python 3.6
- Anaconda
- PyTorch
gymgym[atari]ppaquette_gym_super_mariofceux
The following steps assume Ubuntu 16.04 LTS. During Anaconda installation, press Enter and Yes to accept the defaults.
wget https://repo.continuum.io/archive/Anaconda3-4.4.0-Linux-x86_64.sh
bash Anaconda3-4.4.0-Linux-x86_64.sh
source .bashrc
conda install pytorch torchvision -c soumith
conda install libgcc
pip install gym[Atari]
sudo apt-get update
sudo apt-get install -y python-numpy python-dev cmake zlib1g-dev libjpeg-dev xvfb libav-tools xorg-dev python-opengl libboost-all-dev libsdl2-dev swig
sudo apt-get install fceux
pip install git+https://github.com/ppaquette/gym-super-mario/
- Super Mario Bros
xvfb-run -s "-screen 0 1400x900x24" python train_mario.py
- Atari Pong
python train_pong.py
- Super Mario Bros
Trained on GCP for 16 hours using 8 CPUs. 24 GB of RAM is more than sufficient, but training is hard to converge and it does not consistently clear levels. Training videos are saved by default to /video/mario/.
- Atari Pong
Trained for 8 hours on GCP with 1 GPU (Nvidia Tesla K80) and 4 CPUs. It can reliably and decisively beat the computer. Training videos are saved by default to /video/gym-reslults/.
