Asynchronous deep reinforcement learning
An attempt to repdroduce Google Deep Mind's paper "Asynchronous Methods for Deep Reinforcement Learning."
http:https://arxiv.org/abs/1602.01783
Asynchronous Advantage Actor-Critic (A3C) method for playing "Atari Pong" is implemented with TensorFlow. Both A3C-FF and A3C-LSTM are implemented.
Learning result movment after 26 hours (A3C-FF) is like this.
Any advice or suggestion is strongly welcomed in issues thread.
First we need to build multi thread ready version of Arcade Learning Enviroment. I made some modification to it to run it on multi thread enviroment.
$ git clone https://github.com/miyosuda/Arcade-Learning-Environment.git
$ cd Arcade-Learning-Environment
$ cmake -DUSE_SDL=ON -DUSE_RLGLUE=OFF -DBUILD_EXAMPLES=OFF .
$ make -j 4
$ pip install .
I recommend to install it on VirtualEnv environment.
To train,
$python a3c.py
To display the result with game play,
$python a3c_disp.py
To enable gpu, change "USE_GPU" flag in "constants.py".
When running with 8 parallel game environemts, speeds of GPU (GTX980Ti) and CPU(Core i7 6700) were like this. (Recorded with LOCAL_T_MAX=20 setting.)
type | A3C-FF | A3C-LSTM |
---|---|---|
GPU | 1722 steps per sec | 864 steps per sec |
CPU | 1077 steps per sec | 540 steps per sec |
Score plots of local threads of pong were like these. (with GTX980Ti)
Scores are not averaged using global network unlike the original paper.
- TensorFlow r1.0
- numpy
- cv2
- matplotlib
This project uses setting written in muupan's wiki [muuupan/async-rl] (https://github.com/muupan/async-rl/wiki)
- @aravindsrinivas for providing information for some of the hyper parameters.