parameters.py

class Parameters():

    def __init__(self, env, args):
        # MODEL HYPERPARAMETERS
        # Our input is a stack of 4 frames hence 110x84x4 (Width, height, channels)
        self.state_size = [110, 84, 4]
        self.nb_states = 110 * 84 * 4
        self.nb_actions = env.action_space.n  # 8 possible actions
        self.learning_rate = 0.00025          # Alpha (aka learning rate)

        # TRAINING HYPERPARAMETERS
        self.total_episodes = 500           # Total episodes for training
        self.max_steps = 50000              # Max possible steps in an episode
        self.batch_size = 64                # Batch size

        # Exploration parameters for epsilon greedy strategy
        self.explore_start = 1.0            # exploration probability at start
        self.explore_stop = 0.1             # minimum exploration probability
        self.decay_rate = 0.000003          # exponential decay rate for exploration prob

        # Q learning hyperparameters
        self.gamma = 0.9                    # Discounting rate

        # MEMORY HYPERPARAMETERS
        # Number of experiences stored in the Memory when initialized for the first time
        self.pretrain_length = self.batch_size
        self.memory_size = 20000          # Number of experiences the Memory can keep

        # PREPROCESSING HYPERPARAMETERS
        self.stack_size = 4                 # Number of frames stacked

        # Fixed Q-target : update the parameter of our target_network every tau
        self.tau = 10

        training = bool(int(args[1]))
        episode_render = bool(int(args[2]))
        get_saved_model = bool(int(args[3]))
        simple_dqn = bool(int(args[4]))

        # training = True
        # episode_render = False
        # get_saved_model = False
        # simple_dqn = True

        # MODIFY THIS TO FALSE IF YOU JUST WANT TO SEE THE TRAINED AGENT
        self.training = training

        # TURN THIS TO TRUE IF YOU WANT TO RENDER THE ENVIRONMENT
        self.episode_render = episode_render

        # GET SAVED MODEL (FALSE FOR STARTING WITH NEW FRESH MODELS)
        self.get_saved_model = get_saved_model

        self.simple_dqn = simple_dqn