b7db501d7e0b8404f13be86ed6bcc96f2da5861f,slm_lab/agent/memory/replay.py,Replay,__init__,#Replay#Any#,32

Before Change

        self.actions = np.zeros((self.max_size, self.action_dim))
        self.rewards = np.zeros((self.max_size, 1))
        self.next_states = np.zeros((self.max_size, self.state_dim))
        self.dones = np.zeros((self.max_size, 1))
        self.priorities = np.zeros((self.max_size, 1))

        self.true_size = 0
        self.head = -1  // Index of most recent experience

After Change

        self.action_dim = self.body.action_dim
        self.batch_idxs = None
        self.total_experiences = 0  // to know total size even with forgetting
        self.reset()

    def reset(self):
        self.states = np.zeros((self.max_size, self.state_dim))
        self.actions = np.zeros((self.max_size, self.action_dim))

Instances