I am trying to train Chess data through CNN. To proceed reinforcement learning, I had divided into two - "current network" and "reinforcement network". For each checkpoint file stored in different directory, would like to call the weight and proceed chess games and learn only "reinforcement network."
Thus I had made an instance for each network, current and reinforcement, however, regardless of its sequence, if a network stores saver.restore the other network returns error while same saver.restore process.
I doubt this happens because I had used same variable_scope.
Maybe this is because in tensorflow there already exists same scope_variable I guess.
Any chance I could use two different already-learned network simultaneously on tensorflow?
I use python/tensorflow1.1.0.
Any advice will be appreicated