This is a tensorflow implementation of Asynchronous advantage actor-critic algorithm for CNN-LSTM as function approximator
Training on Breakout-v0 is done with Nvidia GeForce GTX 1070 GPU for 28 hours
- python 3.5
- tensorflow 1.1.0
- opencv 3.2.0
- openAI
For Training Run:
$ python3 trainer.py
For Demo Run:
$ python3 play.py
Got important help form this project