darren1231 / DQN-play-games

Deep Q network github

For quick docker install and train

Make sure install Nvidia Container Toolkit first

• git clone https://github.com/darren1231/DQN-play-games
• cd DQN-play-games/
• docker build -t tf270 .
• docker run -it --gpus all -v /home/darren/code_project/DQN-play-games/:/home/ tf270 /bin/bash

Inside container

Before trainning, uncomment the code in pong_fun.py, owing to pygame monitor can't be shown in docker

## uncomment if train in docker
# os.environ["SDL_VIDEODRIVER"] = "dummy"

• cd /home
• python deep_q_network_normal.py

Show pygame env

After trainning about 400000 steps,we can show the learning effect with pygame monitor without GPU.

Install environment from Anaconda

• conda create -n pygameenv python=3.6
• pip install -r requirements.txt

Before trainning

After trainning