marctuscher/drqn-tensorflow

Deep recurrent Q Learning using Tensorflow, openai/gym and openai/retro

drqn-tensorflow's Issues

Bootstrapped Sequential Updates

Do you try to train model by Bootstrapped Sequential Updates? I do not know how to choose the batch and I will appreciate it if you can help me.

About drqn

Thanks for your code, but I meet some problems.I ran the code used drqn but I found it ran much slower than dqn and the result was also worse than dqn, did I make some mistakes?

requirements.txt

I run the command: pip3 install -r requirements.txt
lots of packages go back to elder version....
I believe you can make it work better.
thanks to you code~

min_history and state_to_update

Hi, I want to use DRQN for my research, and I don't understand the parameters "min_history" and "state_to_update" in train. Could you give me a further explanation?

cannot restore session correctly.

Hello:

I have saved the session and I am trying to restore it. However, I think the reply buffer is referenced before it is defined.

Thanks