Implementations of popular reinfocment learning algorithms integrated with OpenAI Gym. Meant to be easy to understand.
Install the required libraries. First do:
pip install -r requirements.txt
Then you will need to install another library to run atari-py. A forked version that includes Windows support is used. To install it, run the command:
pip install --no-index -f https://github.com/Kojoley/atari-py/releases atari_py
To get started first clone the repo locally. Currently the algorithms implemented are A2C and PPO. To run one of there simply go into its respective folder and run the main.py file.
python main.py
The agent should begin to train automatically and a non training agent will render the game with an up to date network.
To get a live visualization of the agent perfomrance use:
tensorboard --logdir=tb
While inside the same directory as the main.py
Flags are not yet implemented, to change the environment or number of parallel agents simply change the corresponding variables in the main.py file.
- Pablo Beltran - Sturdyplum
- Alex Coleman - alexrcoleman
This project is licensed under the MIT License - see the LICENSE file for details