linecode/rl_atari_pytorch

ReinforcementLearning Learn Play Atari Using DDPG and LSTM.

Atari Pytorch

A atari AI Player implement by pytorch play games

Synopsis

Reinforcement learning shows the most potential of AI in many area, however, to use reinforcement learning you must specific your environment which is somethings hard to build a environment for your problem. But gym let us has a very convenient way to explore rl algorithms.

So here it is, using DDPG and LSTM to play atari, and it is really effective!!, as I can show in Pong-V0-moniter you can find the play progress in mp4. Our AI can really beat computer!!

How to Play With

OK, to play with it, simply run:

./run_train.sh

This will train on Pong-V0 env, and save your model into checkpoints/. If you interrupted, next time it will continue train on last saved model.

And, to play with your model, simply run:

./run_play.sh

You can change env in .sh command, many atari env are supported.

Future

This is a very good exploration but not the end, later on I will explore on reinforcement learning on autonamous-car driving problem and train a AI to fucking drive!!

Contribute

Well, very welcome to send PR to add more game env train models to this repo!! If you have any question about this you can find me via wechat: jintianiloveu.