This project include some state-of-art or classic RL(reinforcement learning) algorithms used for training agents by interactive with Unity through ml-agents v0.8.1.
The Algorithms in this repository are writed totally separated, 'cause I want each algorithm being different with others, what I mean is that I just wanna each algorithm has its own .py file and don't have to switch to another file to find out the implementation which one may confused. And those algorithms will never be encapsulated into a base algorithm model.
This framework implements training mechanism conversion between On-Policy and Off-Policy for Actor-Critic architecture algorithms. Just need to set the value of varibable use_replay_buffer in config_file.py(True for off-policy and False for on-policy).
You can just run each algorithm in this repository by python simple_run.py. I don't put any record function in it(like excel, mongo, logger, checkpoint, summary...).
I am very appreciate to my best friend - BlueFisher - who always look down on my coding style and competence(Although he is right, at least for now, but will be was).
Any questions about this project or errors about my bad grammer, plz let me know in this.