Deep-Reinforcement-Learning-Algorithms-with-PyTorch
Deep-Reinforcement-Learning-Algorithms-with-PyTorch copied to clipboard
p-christ
Issue running h_DQN
Hi,
There are some issues that I encounter when running the h_DQN agent.
Firstly, this line of code would not work if there is no "CONTROLLER" key in the dictionary.
self.controller_config.hyperparameters = self.controller_config.hyperparameters["CONTROLLER"]
I fixed it by add at a key for CONTROLLER and META_CONTROLLER each of which contains the dictionary of config itself.
self.controller.q_network_local = self.create_NN(input_dim=self.state_size*2, output_dim=self.action_size, key_to_use="CONTROLLER")
Secondly, self.controller.q_network_target needs to be created again right after creating the q_network_local to ensure its input dimension is override with self.state_size*2
Please have a review of this and make change to the repo if necessary.
Many thanks for the code! Cheers Khuong
