This particular Agent class is adopted from a Keras example. My goal was to generalize it and make it more production-ready. This particular rendition has the ability to save weights, train on any arbitrary Atari game, and also run through a single episode once training is complete for demo purposes.
The first step is propping up a Virtual Environment that has all of the requirements you need. Before running these steps, make sure you have python3
and virtualenv
installed. To get everything you need to run the DQN Agent, run through the following steps on a Mac:
$ brew install cmake openmpi
$ cd dqn_agent
$ virtualenv ./breakout_env --python=python3
$ source ./breakout_env/bin/activate
$ pip install -r requirements.txt
$ git clone https://github.com/openai/baselines.git
$ cd baselines
$ pip install -e .
$ cd ..
An example of both training and running this agent is provided in breakout.py
.