devsisters / DQN-tensorflow

Tensorflow implementation of Human-Level Control through Deep Reinforcement Learning
MIT License
2.46k stars 765 forks source link

BUG-FIX: Wrong ordering of screen_width and screen_height in tf placeholder #11

Closed pemami4911 closed 7 years ago

pemami4911 commented 7 years ago

In agent.py, the training and target Q networks had the wrong ordering of screen_width and screen_height in the placeholder for s_t. This affects training environments where the resolution has screen_width != screen_height (such as VizDoom)

carpedm20 commented 7 years ago

Thanks for the fix!