I have a bunch of questions that I wanna ask regarding the usage of this library('keras-rl') while using it outside the "Gym" environment. I understand that there are very few users of this library so I may accept a better alternative library.
I am trying to implement a DQN with that T-Rex game on Chrome (chrome://Dino). I was using this library to make an agent when I ran in some problems:-
In my DQN agent the code something like this:
BTW I am using Tensorflow + keras
dqn = DQNAgent(model = model, nb_actions = num_actions,
memory = memory, nb_steps_warmup = 10,target_model_update = 1e-2, policy = strategy)
- So, Since I want to define my own actions (i.e. what controls the agent has) How do I define them?
- In a simple environment with only 2 actions and not much experience what is the recommended memory size?
- In this whole scenario how do I Change my rewards system? Supposing I want to give a reward to the agent for some other thing that what is set by default. How do I do that?
Is there any way to customize the code to change to my requirements? Or is there any other simple to use library to make a DQN agent with minimal setup?
