Method update_state

rl3/a2c/a2c.py:119–122

(self, obs)

        self.real_total_rewards = []

    def update_state(self, obs):
        # Do frame-stacking here instead of the FrameStack wrapper to reduce IPC overhead
        self.state = np.roll(self.state, shift=-self.nc, axis=3)
        self.state[:, :, :, -self.nc:] = obs

    def run(self):
        mb_states, mb_rewards, mb_actions, mb_values, mb_dones = [], [], [], [], []