A Deep Reinforcement Algorithm based on a portability approach of their compositionality present on an environment specific condition.
[sudo] conda create -n bc python=3.6
The creation based on their environment can be a choice made by your prefered tool.
[sudo] [-H] pip install tensorflow gym gym[atari] numpy
or just simple using:
[sudo] pip install -r requirements.txt
Unreach a potential of this tool just utilizing the first capabilitie of then by a choice compounded on their arguments:
[sudo] python app.py
Leveraging from help of their features presented on the app:
[sudo] python app.py --usage help
Basic example with one different game then the default:
python app.py --mode 'render' --environment 'SpaceInvaders-ram-v0' --state_size_environment 'space' --episodes 60 --timesteps 6000 --reinforce 50 --epochs 10 --batch_size 256
All made by feature or fix review, totally welcome from code.