def act()

in robosumo/policy_zoo/policy.py [0:0]

• 14 lines of code
• 2 McCabe index (conditional complexity)

    def act(self, observation, stochastic=True):
        outputs = [self.sampled_action, self.vpred, self.state_out]
        feed_dict = {
            self.observation_ph: observation[None, None],
            self.state_in_ph: list(self.state[:, None, :]),
            self.stochastic_ph: stochastic,
        }
        a, v, s = tf.get_default_session().run(outputs, feed_dict)
        self.state = []
        for x in s:
            self.state.append(x.c[0])
            self.state.append(x.h[0])
        self.state = np.array(self.state)
        return a[0, 0], {'vpred': v[0, 0], 'state': self.state}