simoninithomas

simoninithomas

with tf.Session() as sess:
    
    game, possible_actions = create_environment()
    
    totalScore = 0
    
    # Load the model
    saver.restore(sess, "./models/model.ckpt")
    game.init()
    for i in range(1):

simoninithomas

simoninithomas

class DDDQNNet:
    def __init__(self, state_size, action_size, learning_rate, name):
        self.state_size = state_size
        self.action_size = action_size
        self.learning_rate = learning_rate
        self.name = name
        
        
        # We use tf.variable_scope here to know which network we're using (DQN or target_net)
        # it will be useful when we will update our w- parameters (by copy the DQN parameters)

simoninithomas

### DOUBLE DQN Logic
                # Use DQNNetwork to select the action to take at next_state (a') (action with the highest Q-value)
                # Use TargetNetwork to calculate the Q_val of Q(s',a')
                
                # Get Q values for next_state 
                Qs_next_state = sess.run(DQNetwork.output, feed_dict = {DQNetwork.inputs_: next_states_mb})
                
                # Calculate Qtarget(s')
                q_target = sess.run(TargetNetwork.output, feed_dict = {TargetNetwork.inputs_: next_states_mb})
                

simoninithomas

# Get Q values for next_state 
Qs_next_state = sess.run(TargetNetwork.output, feed_dict = {TargetNetwork.inputs_: next_states_mb})

...

                if tau > max_tau:
                    # Update the parameters of our TargetNetwork with DQN_weights
                    update_target = update_target_graph()
                    sess.run(update_target)
                    tau = 0

simoninithomas

# Instantiate the DQNetwork
DQNetwork = DQNNet(state_size, action_size, learning_rate, name="DQNetwork")

# Instantiate the target network
TargetNetwork = DQNNet(state_size, action_size, learning_rate, name="TargetNetwork")

simoninithomas

# This function helps us to copy one set of variables to another
# In our case we use it when we want to copy the parameters of DQN to Target_network
# Thanks of the very good implementation of Arthur Juliani https://github.com/awjuliani
def update_target_graph():
    
    # Get the parameters of our DQNNetwork
    from_vars = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, "DQNetwork")
    
    # Get the parameters of our Target_network
    to_vars = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, "TargetNetwork")

simoninithomas

simoninithomas