c565f7c9035308fda5e1850e0b9ffbb823cd36ff,chainerrl/agents/soft_actor_critic.py,SoftActorCritic,select_greedy_action,#SoftActorCritic#Any#Any#,293

Before Change

        self.sync_target_network()

    def select_greedy_action(self, obs, deterministic=False):
        with chainer.no_backprop_mode(), chainer.using_config("train", False):
            s = self.batch_states([obs], self.xp, self.phi)
            if deterministic:
                action = self.policy(s).most_probable.array

After Change

        return list(cuda.to_cpu(batch_action))

    def select_greedy_action(self, obs, deterministic=False):
        return self.batch_select_greedy_action(
            [obs], deterministic=deterministic)[0]

    def act_and_train(self, obs, reward):

        self.logger.debug("t:%s r:%s", self.t, reward)

Instances