How to use the tianshou.exploration.GaussianNoise function in tianshou

net = Net(args.layer_num, args.state_shape,
              args.action_shape, concat=True, device=args.device)
    critic = Critic(net, args.device).to(args.device)
    critic_optim = torch.optim.Adam(critic.parameters(), lr=args.critic_lr)
    policy = DDPGPolicy(
        actor, actor_optim, critic, critic_optim,
        args.tau, args.gamma, GaussianNoise(sigma=args.exploration_noise),
        [env.action_space.low[0], env.action_space.high[0]],
        reward_normalization=args.rew_norm,
        ignore_done=args.ignore_done,
        estimation_step=args.n_step)
    # collector
    train_collector = Collector(
        policy, train_envs, ReplayBuffer(args.buffer_size))
    test_collector = Collector(
        policy, test_envs, action_noise=GaussianNoise(sigma=args.test_noise))
    # log
    log_path = os.path.join(args.logdir, args.task, 'ddpg')
    writer = SummaryWriter(log_path)

    def save_fn(policy):
        torch.save(policy.state_dict(), os.path.join(log_path, 'policy.pth'))

    def stop_fn(x):
        return x >= env.spec.reward_threshold

    # trainer
    result = offpolicy_trainer(
        policy, train_collector, test_collector, args.epoch,
        args.step_per_epoch, args.collect_per_step, args.test_num,
        args.batch_size, stop_fn=stop_fn, save_fn=save_fn, writer=writer)
    assert stop_fn(result['best_reward'])

train_envs.seed(args.seed)
    test_envs.seed(args.seed)
    # model
    net = Net(args.layer_num, args.state_shape, device=args.device)
    actor = Actor(
        net, args.action_shape,
        args.max_action, args.device
    ).to(args.device)
    actor_optim = torch.optim.Adam(actor.parameters(), lr=args.actor_lr)
    net = Net(args.layer_num, args.state_shape,
              args.action_shape, concat=True, device=args.device)
    critic = Critic(net, args.device).to(args.device)
    critic_optim = torch.optim.Adam(critic.parameters(), lr=args.critic_lr)
    policy = DDPGPolicy(
        actor, actor_optim, critic, critic_optim,
        args.tau, args.gamma, GaussianNoise(sigma=args.exploration_noise),
        [env.action_space.low[0], env.action_space.high[0]],
        reward_normalization=args.rew_norm,
        ignore_done=args.ignore_done,
        estimation_step=args.n_step)
    # collector
    train_collector = Collector(
        policy, train_envs, ReplayBuffer(args.buffer_size))
    test_collector = Collector(
        policy, test_envs, action_noise=GaussianNoise(sigma=args.test_noise))
    # log
    log_path = os.path.join(args.logdir, args.task, 'ddpg')
    writer = SummaryWriter(log_path)

    def save_fn(policy):
        torch.save(policy.state_dict(), os.path.join(log_path, 'policy.pth'))

# model
    net = Net(args.layer_num, args.state_shape, device=args.device)
    actor = Actor(
        net, args.action_shape,
        args.max_action, args.device
    ).to(args.device)
    actor_optim = torch.optim.Adam(actor.parameters(), lr=args.actor_lr)
    net = Net(args.layer_num, args.state_shape,
              args.action_shape, concat=True, device=args.device)
    critic1 = Critic(net, args.device).to(args.device)
    critic1_optim = torch.optim.Adam(critic1.parameters(), lr=args.critic_lr)
    critic2 = Critic(net, args.device).to(args.device)
    critic2_optim = torch.optim.Adam(critic2.parameters(), lr=args.critic_lr)
    policy = TD3Policy(
        actor, actor_optim, critic1, critic1_optim, critic2, critic2_optim,
        args.tau, args.gamma, GaussianNoise(sigma=args.exploration_noise),
        args.policy_noise, args.update_actor_freq, args.noise_clip,
        [env.action_space.low[0], env.action_space.high[0]],
        reward_normalization=args.rew_norm,
        ignore_done=args.ignore_done,
        estimation_step=args.n_step)
    # collector
    train_collector = Collector(
        policy, train_envs, ReplayBuffer(args.buffer_size))
    test_collector = Collector(policy, test_envs)
    # train_collector.collect(n_step=args.buffer_size)
    # log
    log_path = os.path.join(args.logdir, args.task, 'td3')
    writer = SummaryWriter(log_path)

    def save_fn(policy):
        torch.save(policy.state_dict(), os.path.join(log_path, 'policy.pth'))

np.random.seed(args.seed)
    torch.manual_seed(args.seed)
    train_envs.seed(args.seed)
    test_envs.seed(args.seed)
    # model
    net = Net(args.layer_num, args.state_shape, device=args.device)
    actor = Actor(net, args.action_shape, args.max_action,
                  args.device).to(args.device)
    actor_optim = torch.optim.Adam(actor.parameters(), lr=args.actor_lr)
    net = Net(args.layer_num, args.state_shape,
              args.action_shape, concat=True, device=args.device)
    critic = Critic(net, args.device).to(args.device)
    critic_optim = torch.optim.Adam(critic.parameters(), lr=args.critic_lr)
    policy = DDPGPolicy(
        actor, actor_optim, critic, critic_optim,
        args.tau, args.gamma, GaussianNoise(sigma=args.exploration_noise),
        [env.action_space.low[0], env.action_space.high[0]],
        reward_normalization=True, ignore_done=True)
    # collector
    train_collector = Collector(
        policy, train_envs, ReplayBuffer(args.buffer_size))
    test_collector = Collector(policy, test_envs)
    # log
    writer = SummaryWriter(args.logdir + '/' + 'ddpg')

    def stop_fn(x):
        return x >= env.spec.reward_threshold

    # trainer
    result = offpolicy_trainer(
        policy, train_collector, test_collector, args.epoch,
        args.step_per_epoch, args.collect_per_step, args.test_num,

def __init__(self,
                 actor: torch.nn.Module,
                 actor_optim: torch.optim.Optimizer,
                 critic1: torch.nn.Module,
                 critic1_optim: torch.optim.Optimizer,
                 critic2: torch.nn.Module,
                 critic2_optim: torch.optim.Optimizer,
                 tau: float = 0.005,
                 gamma: float = 0.99,
                 exploration_noise: Optional[BaseNoise]
                 = GaussianNoise(sigma=0.1),
                 policy_noise: float = 0.2,
                 update_actor_freq: int = 2,
                 noise_clip: float = 0.5,
                 action_range: Optional[Tuple[float, float]] = None,
                 reward_normalization: bool = False,
                 ignore_done: bool = False,
                 estimation_step: int = 1,
                 **kwargs) -> None:
        super().__init__(actor, actor_optim, None, None, tau, gamma,
                         exploration_noise, action_range, reward_normalization,
                         ignore_done, estimation_step, **kwargs)
        self.critic1, self.critic1_old = critic1, deepcopy(critic1)
        self.critic1_old.eval()
        self.critic1_optim = critic1_optim
        self.critic2, self.critic2_old = critic2, deepcopy(critic2)
        self.critic2_old.eval()

net = Net(args.layer_num, args.state_shape, device=args.device)
    actor = Actor(
        net, args.action_shape,
        args.max_action, args.device
    ).to(args.device)
    actor_optim = torch.optim.Adam(actor.parameters(), lr=args.actor_lr)
    net = Net(args.layer_num, args.state_shape,
              args.action_shape, concat=True, device=args.device)
    critic1 = Critic(net, args.device).to(args.device)
    critic1_optim = torch.optim.Adam(critic1.parameters(), lr=args.critic_lr)
    critic2 = Critic(net, args.device).to(args.device)
    critic2_optim = torch.optim.Adam(critic2.parameters(), lr=args.critic_lr)
    policy = TD3Policy(
        actor, actor_optim, critic1, critic1_optim, critic2, critic2_optim,
        args.tau, args.gamma,
        GaussianNoise(sigma=args.exploration_noise), args.policy_noise,
        args.update_actor_freq, args.noise_clip,
        [env.action_space.low[0], env.action_space.high[0]],
        reward_normalization=True, ignore_done=True)
    # collector
    train_collector = Collector(
        policy, train_envs, ReplayBuffer(args.buffer_size))
    test_collector = Collector(policy, test_envs)
    # train_collector.collect(n_step=args.buffer_size)
    # log
    writer = SummaryWriter(args.logdir + '/' + 'td3')

    def stop_fn(x):
        if env.spec.reward_threshold:
            return x >= env.spec.reward_threshold
        else:
            return False

net = Net(args.layer_num, args.state_shape, device=args.device)
    actor = Actor(
        net, args.action_shape,
        args.max_action, args.device
    ).to(args.device)
    actor_optim = torch.optim.Adam(actor.parameters(), lr=args.actor_lr)
    net = Net(args.layer_num, args.state_shape,
              args.action_shape, concat=True, device=args.device)
    critic1 = Critic(net, args.device).to(args.device)
    critic1_optim = torch.optim.Adam(critic1.parameters(), lr=args.critic_lr)
    critic2 = Critic(net, args.device).to(args.device)
    critic2_optim = torch.optim.Adam(critic2.parameters(), lr=args.critic_lr)
    policy = TD3Policy(
        actor, actor_optim, critic1, critic1_optim, critic2, critic2_optim,
        args.tau, args.gamma,
        GaussianNoise(sigma=args.exploration_noise), args.policy_noise,
        args.update_actor_freq, args.noise_clip,
        [env.action_space.low[0], env.action_space.high[0]],
        reward_normalization=True, ignore_done=True)
    # collector
    train_collector = Collector(
        policy, train_envs, ReplayBuffer(args.buffer_size))
    test_collector = Collector(policy, test_envs)
    # train_collector.collect(n_step=args.buffer_size)
    # log
    writer = SummaryWriter(args.logdir + '/' + 'td3')

    def stop_fn(x):
        return x >= env.spec.reward_threshold

    # trainer
    result = offpolicy_trainer(