ed592270c7bb5c10e02474d4134ac2b05e7c8b67,chainerrl/envs/abc.py,ABC,step,#ABC#Any#,65

Before Change

            action = np.clip(action, self.action_space.low, self.action_space.high)
            action = np.around(action)
            if action.size > 1:
                action = action[0]
        if action == self._state:
            // Correct
            self._state += 1
            reward = 1.0 if self._state == self.size else 0.0

After Change

            if self.deterministic:
                action = np.argmax(action)
            else:
                pvals = np.exp(action) / np.exp(action).sum()
                action = np.random.multinomial(, pvals)[0]
        if action == self._state:
            // Correct
            self._state += 1

Instances