search

alsora / deep-briscola

Tensorflow Deep Reinforcement Learning agents playing Briscola card game
21 stars 7 forks source link

small fixes to use loaded model #15

Closed alsora closed 5 years ago

alsora commented 5 years ago

After loading a model, you can't use anymore self.events_lenght or similar variables since they are storing old versions of the tensors.

Now self_train works properly. Before I tested it only with dqn where this error was not present.

Plus small fixes