2051aa89ec691d1c056e3e1f1187c489597c2ead,slm_lab/agent/memory/replay.py,Replay,reset,#Replay#,58

Before Change

        """Initializes the memory arrays, size and head pointer"""
        states_shape = np.concatenate([[self.max_size], np.reshape(self.body.state_dim, -1)])
        self.data_keys = ["states", "actions", "rewards", "next_states", "dones", "priorities"]
        setattr(self, "states", np.zeros(states_shape))
        // TODO generalize for multi-action
        setattr(self, "actions", np.zeros((self.max_size,), dtype=self.body.action_space.dtype))
        setattr(self, "rewards", np.zeros((self.max_size,)))
        setattr(self, "next_states", np.zeros(states_shape))
        setattr(self, "dones", np.zeros((self.max_size,), dtype=np.uint8))
        setattr(self, "priorities", np.zeros((self.max_size,)))
        self.true_size = 0
        self.head = -1  // Index of most recent experience

After Change

            elif k == "actions":
                setattr(self, k, np.zeros(self.actions_shape, dtype=self.body.action_space.dtype))
            else:
                setattr(self, k, np.zeros(self.scalar_shape))
        self.true_size = 0
        self.head = -1
        self.state_buffer.clear()
        for _ in range(self.state_buffer.maxlen):

Instances