05_CartPole-reinforcement-learning_PER_D3QN

replay.py

def replay(self):
    if self.USE_PER:
        # Sample minibatch from the PER memory
        tree_idx, minibatch  = self.MEMORY.sample(self.batch_size)
    else:
        # Randomly sample minibatch from the deque memory
        minibatch = random.sample(self.memory, min(len(self.memory), self.batch_size))
    '''
    everything stay the same here as before
    '''
    target_old = np.array(target)
    '''
    everything stay the same here as before
    '''
        if self.USE_PER:
            absolute_errors = np.abs(target_old[i]-target[i])
            # Update priority
            self.MEMORY.batch_update(tree_idx, absolute_errors)

    # Train the Neural Network with batches
    self.model.fit(state, target, batch_size=self.batch_size, verbose=0)