search

devsisters / DQN-tensorflow

Tensorflow implementation of Human-Level Control through Deep Reinforcement Learning
MIT License
2.48k stars 763 forks source link

Help！Why the screen outputs images flow of such low quality as well as with no color? #35

Open xihangalpha opened 7 years ago

xihangalpha commented 7 years ago

When I set is_train=False and display=True in the code, the screen outputs the image flows with rather low quality and without color. Why? Could someone help me?

stefanutti commented 7 years ago

I have a similar problem.

I'm using this command to start docker:

docker run -it -p 8888:8888 -p 6006:6006 --name ai -e PASSWORD=ai -v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=unix$DISPLAY --device /dev/dri --device /dev/snd stefanutti/ai:1.0 bash

I'm using:

dqn-x-server-not-working

nolsigan commented 7 years ago

It is probably due to some dependent libraries of openai universe.

I had same problem and managed it with making new virtualenv without universe. I tried uninstalling universe from current env but it didn't work, so I assume it's not universe itself but its dependent library causing the problem.

I also created an issue on gym github repo but there's no answer yet. https://github.com/openai/gym/issues/596

Raym0ndKwan commented 7 years ago

Ran into exactly the same problem on Mac.

Raym0ndKwan commented 7 years ago

By using an older version of atari-py (0.0.21 for my case), the problem seemed to be solved. Not only works for breakout, also other Atari games.

Summoner918 commented 6 years ago

It is the version problem of gym. By using pip install gym==0.10(my case) and modify environment.py : modify: self.env = gym.make(config.env_name) to: self.env = gym.make(config.env_name).unwrapped It is for all games in gym